A Dual-Augmentor Framework for Domain Generalization in 3D Human Pose Estimation

TL;DR

This paper introduces a dual-augmentor framework with meta-optimization to enhance domain generalization in 3D human pose estimation, effectively handling diverse and unseen scenarios without prior target domain information.

Contribution

The paper proposes a novel dual-augmentor framework with differential strategies and meta-optimization, improving generalization in 3D human pose estimation across diverse domains.

Findings

Outperforms existing domain generalization methods on benchmark datasets

Effectively explores out-of-source pose distributions without prior target domain info

Enhances pose estimator robustness through simulated domain shifts

Abstract

3D human pose data collected in controlled laboratory settings present challenges for pose estimators that generalize across diverse scenarios. To address this, domain generalization is employed. Current methodologies in domain generalization for 3D human pose estimation typically utilize adversarial training to generate synthetic poses for training. Nonetheless, these approaches exhibit several limitations. First, the lack of prior information about the target domain complicates the application of suitable augmentation through a single pose augmentor, affecting generalization on target domains. Moreover, adversarial training's discriminator tends to enforce similarity between source and synthesized poses, impeding the exploration of out-of-source distributions. Furthermore, the pose estimator's optimization is not exposed to domain shifts, limiting its overall generalization ability. To address these limitations, we propose a novel framework featuring two pose augmentors: the weak and the strong augmentors. Our framework employs differential strategies for generation and discrimination processes, facilitating the preservation of knowledge related to source poses and the exploration of out-of-source distributions without prior information about target poses. Besides, we leverage meta-optimization to simulate domain shifts in the optimization process of the pose estimator, thereby improving its generalization ability. Our proposed approach significantly outperforms existing methods, as demonstrated through comprehensive experiments on various benchmark datasets.Our code will be released at \url{https://github.com/davidpengucf/DAF-DG}.