Research Article
A Sparse Feature Matching Model Using a Transformer towards Large-View Indoor Visual Localization
Table 1
Evaluation on HPatches image pairs.
| | Method | AUC | #matches | | 3 px | 5 px | 10 px |
| | D2-net | 23.2 | 35.9 | 53.6 | 0.2 K | | SuperGlue | 53.9 | 68.3 | 81.7 | 0.6 K | | COTR | 62.8 | 67.9 | 80.6 | 1.0 K | | Sparse-NCNet | 48.9 | 54.2 | 67.1 | 1.0 K | | LoFTR | 65.9 | 75.6 | 84.6 | 1.0 K | | MS-T | 28.5 | 48.6 | 52.7 | 1.0 K | | MFA-T | 47.7 | 62.8 | 73.9 | 1.0 K | | MSFA-T | 68.5 | 76.9 | 83.5 | 1.0 K |
|
|
