Drag-based Image Editing on DragBench-SR and DragBench-DR User Study images
61.1CP WinFastDrag
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| FastDragModel Type=Finetuning-free, Evaluation Protocol=A/B testing against ContextDrag2025.12 | 61.1 | 22.9 | 63.4 | 20.9 | 61.5 | 13.3 | |
| Inpaint4DragModel Type=Finetuning-free, Evaluation Protocol=A/B testing against ContextDrag2025.12 | 53.5 | 19.1 | 33.5 | 20.1 | 50 | 17 | |
| DiffEditorModel Type=Finetuning-free, Evaluation Protocol=A/B testing against ContextDrag2025.12 | 53 | 27.1 | 58.2 | 32.6 | 39.3 | 13.4 | |
| DragLoRAModel Type=Finetuning-based, Evaluation Protocol=A/B testing against ContextDrag2025.12 | 46.8 | 27 | 45.2 | 24.4 | 45.6 | 29.3 | |
| SDE-DragModel Type=Finetuning-based, Evaluation Protocol=A/B testing against ContextDrag2025.12 | 46.5 | 23.5 | 58.6 | 21.4 | 52.3 | 13.9 | |
| DragDiffusionModel Type=Finetuning-based, Evaluation Protocol=A/B testing against ContextDrag2025.12 | 38.6 | 26.3 | 47.4 | 5.3 | 50.9 | 24.6 | |
| GoodDragModel Type=Finetuning-based, Evaluation Protocol=A/B testing against ContextDrag2025.12 | 35.3 | 29.7 | 37 | 21.3 | 41.5 | 26 |