Image Registration on VIS-XR Set1 scale 1.7-2.0 (lowest resolution level)
0.8Mean ErrorCN+R+LG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CN+R+LGEstimation Method=DLT, Post-processing=VFC, Components=CraquelureNet, Refine module, LightGlue2026.01 | 0.8 | 2.35 | |
| SP+LG-PEstimation Method=DLT, Fine-tuning=on Multi-modal Panel Painting dataset, Post-processing=VFC2026.01 | 0.81 | 2.48 | |
| CN+LGEstimation Method=DLT, Post-processing=VFC, Components=CraquelureNet, LightGlue2026.01 | 0.81 | 2.26 | |
| Roma-MiEstimation Method=Magsac, Post-processing=VFC2026.01 | 0.89 | 3.38 | |
| Roma-MaEstimation Method=Magsac, Post-processing=VFC2026.01 | 1.13 | 4.44 | |
| Aliked+LG-PEstimation Method=DLT, Fine-tuning=on Multi-modal Panel Painting dataset, Post-processing=VFC2026.01 | 1.18 | 3.89 | |
| LoFTR-MiEstimation Method=Magsac, Post-processing=VFC2026.01 | 1.59 | 6.44 | |
| Roma-OEstimation Method=Magsac, Post-processing=VFC2026.01 | 1.85 | 6.75 | |
| LoFTR-OEstimation Method=Magsac, Post-processing=VFC2026.01 | 3 | 11.45 | |
| LoFTR-PEstimation Method=Magsac, Fine-tuning=on Multi-modal Panel Painting dataset, Post-processing=VFC2026.01 | 9.96 | 43.04 |