Sketch-to-Image Generation on Sketchy In the Wild
1.16FIDIP-Adapter
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| IP-AdapterConditioning (Visual/Textual)=G/G, Evaluation Protocol=Fine-tuned2026.02 | 1.16 | 0.603 | 0.788 | 0.757 | 0.601 | 0.673 | |
| LOTS*Conditioning (Visual/Textual)=L/L, Evaluation Protocol=Fine-tuned2026.02 | 1.19 | 0.629 | 0.808 | 0.749 | 0.663 | 0.673 | |
| LOTSConditioning (Visual/Textual)=L/L, Evaluation Protocol=Fine-tuned2026.02 | 1.23 | 0.636 | 0.817 | 0.735 | 0.684 | 0.709 | |
| ControlNetConditioning (Visual/Textual)=G/G, Evaluation Protocol=Fine-tuned2026.02 | 1.24 | 0.635 | 0.799 | 0.763 | 0.649 | 0.565 | |
| ControlNetConditioning (Visual/Textual)=G/G, Evaluation Protocol=Zero-shot2026.02 | 1.37 | 0.62 | 0.786 | 0.71 | 0.54 | 0.616 | |
| Multi-ControlNetConditioning (Visual/Textual)=L/G, Evaluation Protocol=Zero-shot2026.02 | 1.37 | 0.619 | 0.782 | 0.683 | 0.513 | 0.635 | |
| Multi-ControlNetConditioning (Visual/Textual)=L/G, Evaluation Protocol=Fine-tuned2026.02 | 1.37 | 0.626 | 0.793 | 0.734 | 0.604 | 0.535 | |
| AnyControlConditioning (Visual/Textual)=L/G, Evaluation Protocol=Zero-shot2026.02 | 1.41 | 0.602 | 0.792 | 0.697 | 0.568 | 0.538 | |
| SDConditioning (Visual/Textual)=-/G, Evaluation Protocol=Zero-shot2026.02 | 1.46 | 0.614 | 0.759 | 0.703 | 0.438 | 0.645 | |
| SDXLConditioning (Visual/Textual)=-/G, Evaluation Protocol=Zero-shot2026.02 | 1.46 | 0.566 | 0.759 | 0.787 | 0.541 | 0.661 | |
| GLIGENConditioning (Visual/Textual)=-/L, Evaluation Protocol=Zero-shot2026.02 | 1.57 | 0.538 | 0.709 | 0.293 | 0.223 | 0.599 | |
| GLIGENConditioning (Visual/Textual)=-/L, Evaluation Protocol=Fine-tuned2026.02 | 1.58 | 0.566 | 0.732 | 0.322 | 0.317 | 0.512 | |
| T2I-AdapterConditioning (Visual/Textual)=G/G, Evaluation Protocol=Fine-tuned2026.02 | 1.6 | 0.582 | 0.752 | 0.757 | 0.511 | 0.606 | |
| Multi-T2I-AdapterConditioning (Visual/Textual)=L/G, Evaluation Protocol=Fine-tuned2026.02 | 1.72 | 0.57 | 0.739 | 0.741 | 0.481 | 0.599 | |
| T2I-AdapterConditioning (Visual/Textual)=G/G, Evaluation Protocol=Zero-shot2026.02 | 2.54 | 0.535 | 0.725 | 0.638 | 0.522 | 0.495 | |
| Multi-T2I-AdapterConditioning (Visual/Textual)=L/G, Evaluation Protocol=Zero-shot2026.02 | 2.64 | 0.521 | 0.707 | 0.495 | 0.365 | 0.487 | |
| IP-AdapterConditioning (Visual/Textual)=G/G, Evaluation Protocol=Zero-shot2026.02 | 2.75 | 0.539 | 0.708 | 0.406 | 0.319 | 0.714 |