Saved in:
Bibliographic Details
Main Authors: Zhao, Hongshen, Tai, Jingkang, Wu, Yuhang, Zhang, Wenkang, Lan, Xi, Wang, Shangyan, Zhang, Tianyu, Yang, Wankou
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.24006
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910070880600064
author Zhao, Hongshen
Tai, Jingkang
Wu, Yuhang
Zhang, Wenkang
Lan, Xi
Wang, Shangyan
Zhang, Tianyu
Yang, Wankou
author_facet Zhao, Hongshen
Tai, Jingkang
Wu, Yuhang
Zhang, Wenkang
Lan, Xi
Wang, Shangyan
Zhang, Tianyu
Yang, Wankou
contents Underwater Video Object Segmentation (VOS) is essential for marine exploration, yet open-air methods suffer significant degradation due to color distortion, low contrast, and prevalent camouflage. A primary hurdle is the lack of high-quality training data. To bridge this gap, we introduce $\textbf{UW-VOS}$, the first large-scale underwater VOS benchmark comprising 1,431 video sequences across 409 categories with 309,295 mask annotations, constructed via a semi-automatic data engine with rigorous human verification. We further propose $\textbf{SAM-U}$, a parameter-efficient framework that adapts SAM2 to the underwater domain. By inserting lightweight adapters into the image encoder, SAM-U achieves state-of-the-art performance with only $\sim$2$\%$ trainable parameters. Extensive experiments reveal that existing methods experience an average 13-point $\mathcal{J}\&\mathcal{F}$ drop on UW-VOS, while SAM-U effectively bridges this domain gap. Detailed attribute-based analysis further identifies small targets, camouflage, and exit-re-entry as critical bottlenecks, providing a roadmap for future research in robust underwater perception.
format Preprint
id arxiv_https___arxiv_org_abs_2603_24006
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle UW-VOS: A Large-Scale Dataset for Underwater Video Object Segmentation
Zhao, Hongshen
Tai, Jingkang
Wu, Yuhang
Zhang, Wenkang
Lan, Xi
Wang, Shangyan
Zhang, Tianyu
Yang, Wankou
Computer Vision and Pattern Recognition
Underwater Video Object Segmentation (VOS) is essential for marine exploration, yet open-air methods suffer significant degradation due to color distortion, low contrast, and prevalent camouflage. A primary hurdle is the lack of high-quality training data. To bridge this gap, we introduce $\textbf{UW-VOS}$, the first large-scale underwater VOS benchmark comprising 1,431 video sequences across 409 categories with 309,295 mask annotations, constructed via a semi-automatic data engine with rigorous human verification. We further propose $\textbf{SAM-U}$, a parameter-efficient framework that adapts SAM2 to the underwater domain. By inserting lightweight adapters into the image encoder, SAM-U achieves state-of-the-art performance with only $\sim$2$\%$ trainable parameters. Extensive experiments reveal that existing methods experience an average 13-point $\mathcal{J}\&\mathcal{F}$ drop on UW-VOS, while SAM-U effectively bridges this domain gap. Detailed attribute-based analysis further identifies small targets, camouflage, and exit-re-entry as critical bottlenecks, providing a roadmap for future research in robust underwater perception.
title UW-VOS: A Large-Scale Dataset for Underwater Video Object Segmentation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.24006