Fig. 13From: Retinal vessel segmentation with constrained-based nonnegative matrix factorization and 3D modified attention U-NetVessel segmentation results on HRF dataset, from up to down is input images, ground truth, Soares et al. [13], U-Net, AG-UNet, IterNet, DenseNet, V-GAN, M-GAN, and ours, respectivelyBack to article page