SA-UNetv2: Rethinking Spatial Attention U-Net for Retinal Vessel Segmentation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Guo, Changlu, Christensen, Anders Nymark, Dahl, Anders Bjorholm, Yi, Yugen, Hannemose, Morten Rieger
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914115881009152
author Guo, Changlu
Christensen, Anders Nymark
Dahl, Anders Bjorholm
Yi, Yugen
Hannemose, Morten Rieger
author_facet Guo, Changlu
Christensen, Anders Nymark
Dahl, Anders Bjorholm
Yi, Yugen
Hannemose, Morten Rieger
contents Retinal vessel segmentation is essential for early diagnosis of diseases such as diabetic retinopathy, hypertension, and neurodegenerative disorders. Although SA-UNet introduces spatial attention in the bottleneck, it underuses attention in skip connections and does not address the severe foreground-background imbalance. We propose SA-UNetv2, a lightweight model that injects cross-scale spatial attention into all skip connections to strengthen multi-scale feature fusion and adopts a weighted Binary Cross-Entropy (BCE) plus Matthews Correlation Coefficient (MCC) loss to improve robustness to class imbalance. On the public DRIVE and STARE datasets, SA-UNetv2 achieves state-of-the-art performance with only 1.2MB memory and 0.26M parameters (less than 50% of SA-UNet), and 1 second CPU inference on 592 x 592 x 3 images, demonstrating strong efficiency and deployability in resource-constrained, CPU-only settings.
format Preprint
id arxiv_https___arxiv_org_abs_2509_11774
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SA-UNetv2: Rethinking Spatial Attention U-Net for Retinal Vessel Segmentation
Guo, Changlu
Christensen, Anders Nymark
Dahl, Anders Bjorholm
Yi, Yugen
Hannemose, Morten Rieger
Computer Vision and Pattern Recognition
Retinal vessel segmentation is essential for early diagnosis of diseases such as diabetic retinopathy, hypertension, and neurodegenerative disorders. Although SA-UNet introduces spatial attention in the bottleneck, it underuses attention in skip connections and does not address the severe foreground-background imbalance. We propose SA-UNetv2, a lightweight model that injects cross-scale spatial attention into all skip connections to strengthen multi-scale feature fusion and adopts a weighted Binary Cross-Entropy (BCE) plus Matthews Correlation Coefficient (MCC) loss to improve robustness to class imbalance. On the public DRIVE and STARE datasets, SA-UNetv2 achieves state-of-the-art performance with only 1.2MB memory and 0.26M parameters (less than 50% of SA-UNet), and 1 second CPU inference on 592 x 592 x 3 images, demonstrating strong efficiency and deployability in resource-constrained, CPU-only settings.
title SA-UNetv2: Rethinking Spatial Attention U-Net for Retinal Vessel Segmentation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.11774