Super-Resolution on CelebA (PSNR, SSIM, LPIPS, FID)
29.58PSNRResample
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ResampleNFEs=100, Noise Level (sigma)=0.052024.06 | 29.58 | 0.83 | 24.18 | 45.51 | |
| ProjDiffNFEs=100, Noise Level (sigma)=0.052024.06 | 29.49 | 0.83 | 20.86 | 36.87 | |
| DDRMNFEs=100, Noise Level (sigma)=0.052024.06 | 29.2 | 0.82 | 21.92 | 40.14 | |
| DDNM+NFEs=100, Noise Level (sigma)=0.052024.06 | 29.19 | 0.82 | 21.89 | 40.2 | |
| DMPSNFEs=100, Noise Level (sigma)=0.052024.06 | 29.01 | 0.81 | 22.87 | 37.8 | |
| IGDMNFEs=100, Noise Level (sigma)=0.052024.06 | 28.25 | 0.79 | 22.73 | 38.7 | |
| DPSNFEs=1000, Noise Level (sigma)=0.052024.06 | 27.98 | 0.78 | 23.1 | 39.91 | |
| DiffPIRNFEs=100, Noise Level (sigma)=0.052024.06 | 27.87 | 0.73 | 32.33 | 50.18 | |
| RED-diffNFEs=100, Noise Level (sigma)=0.052024.06 | 24.98 | 0.55 | 50.59 | 73.89 | |
| A†yNoise Level (sigma)=0.052024.06 | 23.64 | 0.49 | 68.72 | 147.89 |