SwinIR: Image Restoration Using Swin Transformer

Benchmark Model Rank Results
color-image-denoising-on-kodak24-sigma50SwinIR#4PSNR: 29.79
color-image-denoising-on-urban100-sigma15-1SwinIR#4Average PSNR: 35.13
color-image-denoising-on-urban100-sigma25SwinIR#4PSNR: 32.9
color-image-denoising-on-urban100-sigma50SwinIR#5PSNR: 29.82
grayscale-image-denoising-on-bsd68-sigma15SwinIR#3PSNR: 31.97
grayscale-image-denoising-on-urban100-sigma15SwinIR#2PSNR: 33.70
grayscale-image-denoising-on-urban100-sigma25SwinIR#4PSNR: 31.3
grayscale-image-denoising-on-urban100-sigma50SwinIR#4PSNR: 27.98
image-super-resolution-on-manga109-4xSwinIR#14SSIM: 0.9273PSNR: 32.22
image-super-resolution-on-set14-4x-upscalingSwinIR#18PSNR: 29.15SSIM: 0.7958
image-super-resolution-on-urban100-4xSwinIR#12PSNR: 27.45SSIM: 0.8254
video-super-resolution-on-msu-super-1SwinIR + x264#2BSQ-rate over Subjective Score: 0.304BSQ-rate over ERQA: 0.76
video-super-resolution-on-msu-super-1SwinIR + x265#4BSQ-rate over Subjective Score: 0.346BSQ-rate over ERQA: 1.575
video-super-resolution-on-msu-super-1SwinIR + uavs3e#8BSQ-rate over Subjective Score: 0.639BSQ-rate over ERQA: 6.803
video-super-resolution-on-msu-super-1SwinIR + aomenc#12BSQ-rate over Subjective Score: 0.835BSQ-rate over ERQA: 10.854
video-super-resolution-on-msu-super-1SwinIR + vvenc#17BSQ-rate over Subjective Score: 1.35BSQ-rate over ERQA: 6.624
video-super-resolution-on-msu-video-upscalersSwinIR-Real-B#3LPIPS: 0.183SSIM: 0.830PSNR: 28.86
video-super-resolution-on-msu-video-upscalersSwinIR-Real-S#5LPIPS: 0.189SSIM: 0.845PSNR: 28.55
video-super-resolution-on-msu-vsr-benchmarkSwinIR#23Subjective score: 4.799ERQAv1.0: 0.6181 - LPIPS: 0.895SSIM: 0.782