Super-resolution on DIV2K (val)
37.32PSNRSST-L+
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SST-L+scale=×2, #params=20.3, Training dataset=DFLIP, training patch size=96 × 962026.03 | 37.32 | 0.9535 | — | — | |
| SST-Lscale=×2, #params=20.3, Training dataset=DFLIP2026.03 | 37.21 | 0.9528 | — | — | |
| PFTscale=×2, #params=19.6, Training dataset=DFLIP2026.03 | 37.17 | 0.9526 | — | — | |
| MambaIRV2-Lscale=×2, #params=34.2, Training dataset=DFLIP2026.03 | 37.16 | 0.9524 | — | — | |
| SRResNet-GradResidualScaling=sample-based2026.02 | 33.8 | — | — | — | |
| SST-L+scale=×3, #params=20.7, Training dataset=DFLIP, training patch size=96 × 962026.03 | 33.55 | 0.9037 | — | — | |
| SST-Lscale=×3, #params=20.7, Training dataset=DFLIP2026.03 | 33.45 | 0.9026 | — | — | |
| MambaIRV2-Lscale=×3, #params=34.2, Training dataset=DFLIP2026.03 | 33.43 | 0.902 | — | — | |
| PFTscale=×3, #params=19.8, Training dataset=DFLIP2026.03 | 33.41 | 0.902 | — | — | |
| SRResNetScaling=sample-based2026.02 | 33.16 | — | — | — | |
| SST-L+scale=×4, #params=21.3, Training dataset=DFLIP, training patch size=96 × 962026.03 | 31.5 | 0.8594 | — | — | |
| SST-Lscale=×4, #params=21.3, Training dataset=DFLIP2026.03 | 31.44 | 0.8584 | — | — | |
| MambaIRV2-Lscale=×4, #params=34.2, Training dataset=DFLIP2026.03 | 31.4 | 0.8576 | — | — | |
| PFTscale=×4, #params=19.8, Training dataset=DFLIP2026.03 | 31.38 | 0.8573 | — | — | |
| ABPN_FIRS=x32022.08 | 30.18 | — | — | — | |
| ABPNS=x32022.08 | 30.14 | — | — | — | |
| SeDGenerator Backbone=RRDB, Discriminator Architecture=U-Net + SeD2024.02 | 29.85 | 0.818 | 0.102 | — | |
| SeDGenerator Backbone=SwinIR, Discriminator Architecture=U-Net + SeD2024.02 | 29.79 | 0.816 | 0.096 | — | |
| PatchGANGenerator Backbone=SwinIR, Discriminator Architecture=PatchGAN2024.02 | 29.66 | 0.815 | 0.103 | — | |
| U-NetGenerator Backbone=SwinIR, Discriminator Architecture=U-Net2024.02 | 29.56 | 0.81 | 0.095 | — | |
| SeDGenerator Backbone=SwinIR, Discriminator Architecture=PatchGAN + SeD2024.02 | 29.53 | 0.81 | 0.09 | — | |
| RRDBScaling Factor=4x2023.04 | 29.44 | 0.84 | 0.253 | — | |
| DualFormerGenerator Backbone=RRDB2024.02 | 29.3 | 0.802 | 0.103 | — | |
| U-NetGenerator Backbone=RRDB, Discriminator Architecture=U-Net2024.02 | 29.28 | 0.802 | 0.11 | — | |
| SeDGenerator Backbone=RRDB, Discriminator Architecture=PatchGAN + SeD2024.02 | 29.27 | 0.802 | 0.094 | — | |
| LDLGenerator Backbone=SwinIR2024.02 | 29.12 | 0.801 | 0.094 | — | |
| DATPRL-IR-9TMethod Type=Multi-Domain All-in-One Method2026.03 | 29.05 | 0.8181 | — | — | |
| DATPRL-IR-9TYear=(Ours), Method Type=Multi-Domain All-in-One Method2026.03 | 29.05 | 0.8181 | — | — | |
| DATPRL-IR-7TYear=(Ours), Method Type=Multi-Domain All-in-One Method2026.03 | 29.03 | 0.8183 | — | — | |
| LIIFCategory=Reg.-based, Datasets=D+F2023.03 | 29 | 0.89 | — | — | |
| DATPRL-IR-8TYear=(Ours), Method Type=Multi-Domain All-in-One Method2026.03 | 28.99 | 0.8188 | — | — | |
| OFTSR distilledt=0, NFEs=1, Training Protocol=Training-based2024.12 | 28.99 | — | 0.271 | 18.07 | |
| EDSRCategory=Reg.-based, Datasets=D+F2023.03 | 28.98 | 0.83 | — | — | |
| EDSRScaling Factor=4x2023.04 | 28.98 | 0.83 | 0.27 | — | |
| DATPRL-IR-6TYear=(Ours), Method Type=Multi-Domain All-in-One Method2026.03 | 28.98 | 0.8191 | — | — | |
| LDLGenerator Backbone=RRDB2024.02 | 28.95 | 0.795 | 0.101 | — | |
| RestormerYear=CVPR2022, Method Type=Single-Task Method2026.03 | 28.94 | 0.8158 | — | — | |
| PromptIRMethod Type=All-in-One Method2026.03 | 28.86 | 0.8127 | — | — | |
| MPRNetYear=CVPR2021, Method Type=Single-Task Method2026.03 | 28.82 | 0.8115 | — | — | |
| AdaIRYear=ICLR2025, Method Type=All-in-One Method2026.03 | 28.81 | 0.8157 | — | — | |
| SwinIRMethod Type=Single-Task Method2026.03 | 28.8 | 0.8109 | — | — | |
| AMIRYear=MICCAI2024, Method Type=All-in-One Method2026.03 | 28.78 | 0.8139 | — | — | |
| USRGANGenerator Backbone=USRGAN2024.02 | 28.77 | 0.793 | 0.132 | — | |
| PromptIRYear=NeurIPS2023, Method Type=All-in-One Method2026.03 | 28.77 | 0.816 | — | — | |
| NAFNetYear=ECCV2022, Method Type=Single-Task Method2026.03 | 28.73 | 0.8146 | — | — | |
| PatchGANGenerator Backbone=RRDB, Discriminator Architecture=PatchGAN2024.02 | 28.71 | 0.792 | 0.111 | — | |
| MoCEIRMethod Type=All-in-One Method2026.03 | 28.68 | 0.8152 | — | — | |
| NAFNetMethod Type=Single-Task Method2026.03 | 28.64 | 0.8128 | — | — | |
| RestormerMethod Type=Single-Task Method2026.03 | 28.63 | 0.815 | — | — | |
| SwinIRYear=ICCVW2021, Method Type=Single-Task Method2026.03 | 28.61 | 0.8051 | — | — | |
| Mean-ODEUpscaling Factor=4x2023.12 | 28.5 | 0.807 | 0.328 | 22.14 | |
| MPRNetMethod Type=Single-Task Method2026.03 | 28.32 | 0.8067 | — | — | |
| AdaIRMethod Type=All-in-One Method2026.03 | 28.24 | 0.8153 | — | — | |
| ESRGANGenerator Backbone=ESRGAN2024.02 | 28.2 | 0.777 | 0.115 | — | |
| TransweatherMethod Type=All-in-One Method2026.03 | 28.16 | 0.7951 | — | — | |
| MoCEIRYear=CVPR2025, Method Type=All-in-One Method2026.03 | 28.16 | 0.8156 | — | — | |
| DiWaScaling Factor=4x2023.04 | 28.09 | 0.78 | 0.104 | — | |
| DDNMNFEs=100, Training Protocol=Training-free2024.12 | 28.09 | — | 0.279 | 20.33 | |
| SFTGANGenerator Backbone=SFTGAN2024.02 | 28.08 | 0.771 | 0.133 | — | |
| OFTSR distilledt=0.5, NFEs=1, Training Protocol=Training-based2024.12 | 28.02 | — | 0.208 | 16.89 | |
| DiffPIRNFEs=100, Training Protocol=Training-free2024.12 | 27.94 | — | 0.248 | 19.56 | |
| DDRMNFEs=20, Training Protocol=Training-free2024.12 | 27.87 | — | 0.285 | 23.38 | |
| DFPIRYear=CVPR2025, Method Type=All-in-One Method2026.03 | 27.69 | 0.7845 | — | — | |
| IDMCategory=Diffusion, Datasets=D+F2023.03 | 27.59 | 0.78 | — | — | |
| SRDiffScaling Factor=4x2023.04 | 27.41 | 0.79 | 0.136 | — | |
| TransweatherYear=CVPR2022, Method Type=All-in-One Method2026.03 | 27.4 | 0.7643 | — | — | |
| ECDBNFEs=100, Training Protocol=Training-based2024.12 | 27.39 | — | 0.212 | 18.88 | |
| IDMCategory=Diffusion, Datasets=D2023.03 | 27.1 | 0.77 | — | — | |
| SRFlowCategory=Flow-based, Datasets=D+F2023.03 | 27.09 | 0.76 | — | — | |
| SRFlowScaling Factor=4x2023.04 | 27.09 | 0.76 | 0.12 | — | |
| LAR-SRCategory=VAE+AR, Datasets=D+F2023.03 | 27.03 | 0.77 | — | — | |
| HCFlowCategory=Flow-based, Datasets=D+F2023.03 | 27.02 | 0.76 | — | — | |
| GOUBNFEs=100, Training Protocol=Training-based2024.12 | 26.92 | — | 0.218 | 21.56 | |
| GOUBUpscaling Factor=4x2023.12 | 26.89 | 0.7478 | 0.22 | 20.85 | |
| OFTSR distilledt=1, NFEs=1, Training Protocol=Training-based2024.12 | 26.87 | — | 0.127 | 14.58 | |
| IRSDENFEs=100, Training Protocol=Training-based2024.12 | 26.83 | — | 0.144 | 14.69 | |
| OFTSRNFEs=31, Training Protocol=Training-based2024.12 | 26.76 | — | 0.128 | 14.1 | |
| BicubicDatasets=D+F2023.03 | 26.7 | 0.77 | — | — | |
| BicubicScaling Factor=4x2023.04 | 26.7 | 0.77 | 0.409 | — | |
| CNN-BASELINECrop Size=256x2562023.01 | 26.64 | 0.6729 | 0.389 | 133.95 | |
| HCFlow++Category=Flow+GAN, Datasets=D+F2023.03 | 26.61 | 0.74 | — | — | |
| RankSRGANCategory=GAN-based, Datasets=D+F2023.03 | 26.55 | 0.75 | — | — | |
| RankSRGANScaling Factor=4x2023.04 | 26.55 | 0.75 | 0.128 | — | |
| InDINFEs=100, Training Protocol=Training-based2024.12 | 26.45 | — | 0.136 | 15.39 | |
| ESRGANCategory=GAN-based, Datasets=D+F2023.03 | 26.22 | 0.75 | — | — | |
| ESRGANScaling Factor=4x2023.04 | 26.22 | 0.75 | 0.124 | — | |
| IR-SDE (Denoising-ODE)Crop Size=256x2562023.01 | 25.9 | 0.657 | 0.231 | 45.36 | |
| IR-SDEUpscaling Factor=4x2023.12 | 25.9 | 0.657 | 0.231 | 45.36 | |
| DDRMCrop Size=256x2562023.01 | 24.35 | 0.5927 | 0.364 | 78.71 | |
| DDRMUpscaling Factor=4x2023.12 | 24.35 | 0.5927 | 0.364 | 78.71 | |
| DPSNFEs=1000, Training Protocol=Training-free2024.12 | 23.05 | — | 0.447 | 109.35 |