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

Abstract

This paper introduces a novel Dual-Augmentor Framework aimed at enhancing domain generalization in 3D human pose estimation. Domain generalization, the ability of models to perform well on unseen domains, is crucial for deploying pose estimation systems in real-world scenarios characterized by diverse environmental conditions and data sources. Our framework addresses this challenge by leveraging two distinct augmentors during training: one focusing on domain-specific features and the other on domain-agnostic representations. By integrating these augmentors within a unified framework, our method effectively bridges the domain gap, enabling accurate pose estimation across varied environments. Extensive experiments on benchmark datasets demonstrate the efficacy of the Dual-Augmentor Framework, showcasing significant improvements in domain generalization performance compared to existing methods. Overall, our contributions present a promising approach for advancing the robustness and adaptability of 3D human pose estimation systems in real-world applications. This paper presents a novel Dual-Augmentor Framework aimed at enhancing domain generalization in 3D human pose estimation. Addressing the challenge of domain shift, where models trained on data from one domain perform poorly on data from another domain, our approach integrates two distinct augmentors within a unified framework. One augmentor emphasizes domain-specific features, while the other focuses on extracting domain-agnostic representations. By leveraging this dual augmentation strategy, our framework bridges the domain gap, enabling accurate pose estimation across varied environments.