Adversarial Parametric Pose Prior

The Skinned Multi-Person Linear (SMPL) model represents human bodies by mapping pose and shape parameters to body meshes. However, not all pose and shape parameter values yield physically-plausible or even realistic body meshes. In other words, SMPL is under-constrained and may yield invalid results. We propose learning a prior that restricts the SMPL parameters to values that produce realistic poses via adversarial training. We show that our learned prior covers the diversity of the real-data distribution, facilitates optimization for 3D reconstruction from 2D keypoints, and yields better pose estimates when used for regression from images. For all these tasks, it outperforms the state-of-the-art VAE-based approach to constraining the SMPL parameters. The code will be made available at https://github.com/cvlab-epfl/adv_param_pose_prior.

1. Introduction

The SMPL model [17] is now widely used to parameterize body poses and shapes [15], [21]. However, it offers no guarantee to produce realistic human bodies when random values are passed as its inputs. This complicates its usage within an optimization, regression, or generative frameworks, where it is desirable that any sample drawn be plausible.