MoExDA: Domain Adaptation for Edge-based Action Recognition

TL;DR

MoExDA introduces a lightweight domain adaptation technique that leverages edge frames alongside RGB data to mitigate static bias in action recognition models, enhancing robustness with lower computational costs.

Contribution

The paper presents MoExDA, a novel domain adaptation method that effectively reduces static bias in action recognition by integrating edge information, improving robustness and efficiency.

Findings

Effective suppression of static bias in action recognition.

Improved robustness over previous methods.

Lower computational cost achieved.

Abstract

Modern action recognition models suffer from static bias, leading to reduced generalization performance. In this paper, we propose MoExDA, a lightweight domain adaptation between RGB and edge information using edge frames in addition to RGB frames to counter the static bias issue. Experiments demonstrate that the proposed method effectively suppresses static bias with a lower computational cost, allowing for more robust action recognition than previous approaches.