Sequential Image-to-Image Editing on Targeted Task Pairs Deblurring -> Deraining
39.8LPIPST2T-VICL
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| T2T-VICLBase Model=Seedream 4.02025.11 | 39.8 | — | 23.9 | 113.8 | — | |
| Seedream 4.0Variant=Fixed2025.11 | 41.4 | — | 25 | 135.9 | — | |
| T2T-VICLBase Model=Qwen-Image-Edit2025.11 | 65.3 | — | 31.7 | 171.2 | — | |
| Qwen-Image-EditVariant=Fixed2025.11 | 68.9 | — | 34.2 | 178.7 | — | |
| T2T-VICLBase Model=FLUX.2-dev2025.11 | 75 | — | 35.9 | 222.2 | — | |
| FLUX.2-devVariant=Fixed2025.11 | 76.5 | — | 37.3 | 244.7 | — | |
| OmniGen 2Variant=Fixed2025.11 | 78.2 | — | 40.2 | 266.4 | — | |
| T2T-VICLBase Model=OmniGen 22025.11 | 78.2 | — | 39.2 | 249.7 | — |