The author introduces the Swin-Res-Net model to enhance retinal vessel segmentation by combining the Swin Transformer and Res2net, achieving superior performance metrics.
The proposed Graph Capsule Convolution Network (GCC-UNet) effectively captures both local and global features to achieve superior retinal vessel segmentation performance.
A lightweight encoder-decoder network with region-guided attention blocks and weighted dice loss for accurate and efficient segmentation of retinal vessels.