MambaU-Lite: A Lightweight Model based on Mamba and Integrated Channel-Spatial Attention for Skin Lesion Segmentation

Thi-Nhu-Quynh Nguyen, Quang-Huy Ho, Duy-Thai Nguyen, Hoang-Minh-Quang Le, Van-Truong Pham, Thi-Thao Tran · 2024

Read MambaU-Lite: A Lightweight Model based on Mamba and Integrated Channel-Spatial Attention for Skin Lesion Segmentation by Thi-Nhu-Quynh Nguyen, Quang-Huy Ho, Duy-Thai Nguyen, Hoang-Minh-Quang Le, Van-Truong Pham, Thi-Thao Tran (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-01405.txt.

Description

Nguyen and five co-authors introduced a lightweight computer vision model in twenty twenty-four for outlining skin lesions in medical images. Marking the exact edge of a spot helps doctors judge skin cancer early, but the borders are often fuzzy and clinic devices cannot run heavy models.

Their model, MambaU-Lite, blends two design families: the Mamba design, which reads long patterns well, and classic image filters that catch fine local detail. A new block pools features at several scales so the model sees both the whole lesion and its edge. The whole system uses just over four hundred thousand settings, small enough for modest hardware. Tests on two public skin image sets gave promising scores. The authors plan to release the code, which could bring careful lesion checks to smaller clinics.

Catalog record

CREATORThi-Nhu-Quynh Nguyen, Quang-Huy Ho, Duy-Thai Nguyen, Hoang-Minh-Quang Le, Van-Truong Pham, Thi-Thao Tran
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE18 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-01405

Browse the full library · About the Tower of Babel