Skip to main content

Table 1 Comparison of SSIM and PSNR on DRISHTI-GS dataset with different GAN models

From: Retinal image synthesis from multiple-landmarks input with generative adversarial networks

Methods

SSIM

PSNR

Pix2pix

 ResNet-6 vessel

0.8983

23.8254

 ResNet-6 MCML

0.8968

24.0820

 ResNet-9 vessel

0.8914

23.5531

 ResNet-9 MCML

0.9079

25.3665

 U-net vessel

0.8956

23.4331

 U-net MCML

0.9117

24.5544

Cycle-GAN

 ResNet-6 vessel

0.8770

22.4830

 ResNet-6 MCML

0.9019

23.8092

 ResNet-9 vessel

0.8774

21.9402

 ResNet-9 MCML

0.8907

23.1031

 U-net vessel

0.8984

22.7353

 U-net MCML

0.8877

23.0110

  1. Italic values are the best results