Abstract
We present Pose-NDF, a continuous model for plausible human poses based onneural distance fields (NDFs). Pose or motion priors are important forgenerating realistic new poses and for reconstructing accurate poses from noisyor partial observations. Pose-NDF learns a manifold of plausible poses as thezero level set of a neural implicit function, extending the idea of modelingimplicit surfaces in 3D to the high-dimensional domain SO(3)^K, where a humanpose is defined by a single data point, represented by K quaternions. Theresulting high-dimensional implicit function can be differentiated with respectto the input poses and thus can be used to project arbitrary poses onto themanifold by using gradient descent on the set of 3-dimensional hyperspheres. Incontrast to previous VAE-based human pose priors, which transform the posespace into a Gaussian distribution, we model the actual pose manifold,preserving the distances between poses. We demonstrate that PoseNDF outperformsexisting state-of-the-art methods as a prior in various downstream tasks,ranging from denoising real-world human mocap data, pose recovery from occludeddata to 3D pose reconstruction from images. Furthermore, we show that it can beused to generate more diverse poses by random sampling and projection thanVAE-based methods.