GFPose: Learning 3D Human Pose Prior with Gradient Fields

TL;DR

GFPose is a versatile framework that models 3D human pose priors using a gradient-based denoising approach, improving accuracy and diversity in pose estimation and generation tasks.

Contribution

It introduces a time-dependent score network for 3D human pose modeling, unifying discriminative and generative tasks within a single, simple framework.

Findings

Outperforms state-of-the-art in multi-hypothesis pose estimation by 20%.

Achieves comparable results to deterministic methods with a simple backbone.

Capable of generating diverse and realistic 3D human poses.

Abstract

Learning 3D human pose prior is essential to human-centered AI. Here, we present GFPose, a versatile framework to model plausible 3D human poses for various applications. At the core of GFPose is a time-dependent score network, which estimates the gradient on each body joint and progressively denoises the perturbed 3D human pose to match a given task specification. During the denoising process, GFPose implicitly incorporates pose priors in gradients and unifies various discriminative and generative tasks in an elegant framework. Despite the simplicity, GFPose demonstrates great potential in several downstream tasks. Our experiments empirically show that 1) as a multi-hypothesis pose estimator, GFPose outperforms existing SOTAs by 20% on Human3.6M dataset. 2) as a single-hypothesis pose estimator, GFPose achieves comparable results to deterministic SOTAs, even with a vanilla backbone. 3) GFPose is able to produce diverse and realistic samples in pose denoising, completion and generation tasks. Project page https://sites.google.com/view/gfpose/