Enregistré dans:
Détails bibliographiques
Auteurs principaux: Wei, Pengfei, Sun, Yiqun, Xu, Zhiqiang, Ke, Yiping, Hsieh, Lawrence B.
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:https://arxiv.org/abs/2605.19510
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866909056171507712
author Wei, Pengfei
Sun, Yiqun
Xu, Zhiqiang
Ke, Yiping
Hsieh, Lawrence B.
author_facet Wei, Pengfei
Sun, Yiqun
Xu, Zhiqiang
Ke, Yiping
Hsieh, Lawrence B.
contents Unsupervised video domain adaptation (UVDA) is a practical but under-explored problem. In this paper, we propose a frustratingly easy UVDA method, called MetaTrans. Specifically, MetaTrans adopts a concise learning objective that contains only two fundamental loss terms. Despite the simplicity of the learning objective, MetaTrans embodies an advanced UVDA idea, that is, handling the spatial and temporal divergence of cross-domain videos separately, through a subtle model architecture design. By implementing a temporal-static subtraction module, MetaTrans effectively removes spatial and temporal divergence. Extensive empirical evaluations, particularly on various cross-domain action recognition tasks, show substantial absolute adaptation performance enhancement and significantly superior relative performance gain compared with state-of-the-art UVDA baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2605_19510
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Return of Frustratingly Easy Unsupervised Video Domain Adaptation
Wei, Pengfei
Sun, Yiqun
Xu, Zhiqiang
Ke, Yiping
Hsieh, Lawrence B.
Computer Vision and Pattern Recognition
Unsupervised video domain adaptation (UVDA) is a practical but under-explored problem. In this paper, we propose a frustratingly easy UVDA method, called MetaTrans. Specifically, MetaTrans adopts a concise learning objective that contains only two fundamental loss terms. Despite the simplicity of the learning objective, MetaTrans embodies an advanced UVDA idea, that is, handling the spatial and temporal divergence of cross-domain videos separately, through a subtle model architecture design. By implementing a temporal-static subtraction module, MetaTrans effectively removes spatial and temporal divergence. Extensive empirical evaluations, particularly on various cross-domain action recognition tasks, show substantial absolute adaptation performance enhancement and significantly superior relative performance gain compared with state-of-the-art UVDA baselines.
title Return of Frustratingly Easy Unsupervised Video Domain Adaptation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2605.19510