Two-frame relative pose estimation on Mannequin Challenge (MC) (test)
36.24AUC @15°Our SfM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Our SfMcomponents=Virtual Correspondences (VCs) + classic correspondences2022.06 | 36.24 | 61.38 | 73.2 | |
| LoFTRRefinement=Bundle Adjustment (BA)2022.06 | 27.51 | 36.32 | 40.55 | |
| LoFTRpose estimation algorithm=five-point algorithm2022.06 | 27.47 | 35.98 | 40.1 | |
| SuperGluepose estimation algorithm=five-point algorithm2022.06 | 26.38 | 34.85 | 39.1 | |
| SuperGlueRefinement=Bundle Adjustment (BA)2022.06 | 26.24 | 35.12 | 39.46 | |
| SuperPointRefinement=Bundle Adjustment (BA)2022.06 | 17.12 | 23.48 | 26.81 | |
| Deep optimization2022.06 | 15.38 | 47.08 | 63.67 | |
| SIFTRefinement=Bundle Adjustment (BA)2022.06 | 14.17 | 20.24 | 24.25 | |
| Deep regression2022.06 | 4.61 | 11.23 | 16.44 |