Multi-View Camera Pose Estimation on Texture-Poor SfM Dataset
49.94AUC (Threshold 3°)RoMa + Ours
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| RoMa + OursType=Dense Matching, Sub-type=Deep2025.01 | 49.94 | 66.23 | 81.41 | |
| Mast3r + COLMAP + Our RefinementType=Point-Based2025.01 | 48.03 | 64.48 | 80.31 | |
| Mast3r + COLMAPType=Point-Based2025.01 | 43.46 | 61.51 | 78.97 | |
| DKM + OursType=Dense Matching, Sub-type=Deep2025.01 | 41.19 | 57.26 | 72.95 | |
| RoMa + DFSfMType=Dense Matching2025.01 | 34.28 | 50.01 | 69.01 | |
| MatchFormer + DFSfMType=Semi-Dense Matching2025.01 | 26.9 | 37.57 | 48.55 | |
| LOFTR + DFSfMType=Semi-Dense Matching2025.01 | 26.07 | 35.77 | 45.43 | |
| AspanTrans. + DFSfMType=Semi-Dense Matching2025.01 | 25.78 | 35.69 | 45.64 | |
| LOFTR + PixSfMType=Detector-Free2025.01 | 20.66 | 30.49 | 42.01 | |
| SP + SG + PixSfMType=Detector-Based2025.01 | 14 | 19.23 | 24.55 | |
| Mast3r-SfMType=Point-Based2025.01 | 7.1 | 18.08 | 42.44 | |
| R2D2 + NN + PixSfMType=Detector-Based2025.01 | 3.79 | 5.51 | 7.84 | |
| SIFT + NN + PixSfMType=Detector-Based2025.01 | 3.13 | 4.08 | 5.09 | |
| COLMAP (SIFT+NN)Type=Detector-Based2025.01 | 2.87 | 3.85 | 4.95 | |
| VGGSfMType=Dense Matching, Sub-type=Deep2025.01 | 1.86 | 6.26 | 21.1 | |
| D2Net + NN + PixSfMType=Detector-Based2025.01 | 1.54 | 2.63 | 4.54 |