Foreground-Aware Stylization and Consensus Pseudo-Labeling for Domain Adaptation of First-Person Hand Segmentation

TL;DR

This paper introduces a novel domain adaptation method for first-person hand segmentation, combining foreground-aware stylization and consensus pseudo-labeling to improve performance across different environments.

Contribution

It proposes a new approach that stylizes images separately for foreground and background and uses consensus pseudo-labeling to effectively adapt models to new domains.

Findings

Achieved state-of-the-art results in real and simulation domain adaptation.

Effective in multi-target domain adaptation and generalization.

Code availability facilitates reproducibility.

Abstract

Hand segmentation is a crucial task in first-person vision. Since first-person images exhibit strong bias in appearance among different environments, adapting a pre-trained segmentation model to a new domain is required in hand segmentation. Here, we focus on appearance gaps for hand regions and backgrounds separately. We propose (i) foreground-aware image stylization and (ii) consensus pseudo-labeling for domain adaptation of hand segmentation. We stylize source images independently for the foreground and background using target images as style. To resolve the domain shift that the stylization has not addressed, we apply careful pseudo-labeling by taking a consensus between the models trained on the source and stylized source images. We validated our method on domain adaptation of hand segmentation from real and simulation images. Our method achieved state-of-the-art performance in both settings. We also demonstrated promising results in challenging multi-target domain adaptation and domain generalization settings. Code is available at https://github.com/ut-vision/FgSty-CPL.