SAM-DCE: Addressing Token Uniformity and Semantic Over-Smoothing in Medical Segmentation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Hu, Yingzhen, Zhong, Yiheng, Li, Ruobing, Su, Yingxue, An, Jiabao, Tang, Feilong, Su, Jionglong, Razzak, Imran
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915508159250432
author Hu, Yingzhen
Zhong, Yiheng
Li, Ruobing
Su, Yingxue
An, Jiabao
Tang, Feilong
Su, Jionglong
Razzak, Imran
author_facet Hu, Yingzhen
Zhong, Yiheng
Li, Ruobing
Su, Yingxue
An, Jiabao
Tang, Feilong
Su, Jionglong
Razzak, Imran
contents The Segment Anything Model (SAM) demonstrates impressive zero-shot segmentation ability on natural images but encounters difficulties in medical imaging due to domain shifts, anatomical variability, and its reliance on user-provided prompts. Recent prompt-free adaptations alleviate the need for expert intervention, yet still suffer from limited robustness and adaptability, often overlooking the issues of semantic over-smoothing and token uniformity. We propose SAM-DCE, which balances local discrimination and global semantics while mitigating token uniformity, enhancing inter-class separability, and enriching mask decoding with fine-grained, consistent representations. Extensive experiments on diverse medical benchmarks validate its effectiveness.
format Preprint
id arxiv_https___arxiv_org_abs_2509_16886
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SAM-DCE: Addressing Token Uniformity and Semantic Over-Smoothing in Medical Segmentation
Hu, Yingzhen
Zhong, Yiheng
Li, Ruobing
Su, Yingxue
An, Jiabao
Tang, Feilong
Su, Jionglong
Razzak, Imran
Computer Vision and Pattern Recognition
The Segment Anything Model (SAM) demonstrates impressive zero-shot segmentation ability on natural images but encounters difficulties in medical imaging due to domain shifts, anatomical variability, and its reliance on user-provided prompts. Recent prompt-free adaptations alleviate the need for expert intervention, yet still suffer from limited robustness and adaptability, often overlooking the issues of semantic over-smoothing and token uniformity. We propose SAM-DCE, which balances local discrimination and global semantics while mitigating token uniformity, enhancing inter-class separability, and enriching mask decoding with fine-grained, consistent representations. Extensive experiments on diverse medical benchmarks validate its effectiveness.
title SAM-DCE: Addressing Token Uniformity and Semantic Over-Smoothing in Medical Segmentation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.16886