SA Unet Improved

TL;DR

This paper introduces a novel segmentation method combining StyleGAN2 and SA-Unet to improve retinal vessel segmentation, especially effective on small datasets, aiding medical diagnosis.

Contribution

The paper presents a new approach integrating StyleGAN2 with SA-Unet for enhanced retinal vessel segmentation on limited data.

Findings

Effective segmentation on small datasets

Improved accuracy in retinal vessel detection

Potential for broader medical image segmentation

Abstract

Retinal vessels segmentation is well known problem in image processing on the medical field. Good segmentation may help doctors take better decisions while diagnose eyes disuses. This paper describes our work taking up the DRIVE challenge which include segmentation on retinal vessels. We invented a new method which combine using of StyleGAN2 and SA-Unet. Our innovation can help any small data set segmentation problem.