Table 6 Anomaly detection performance of different curvature values \(\rho\) under different datasets.
From: Hyperbolic geometry enhanced feature filtering network for industrial anomaly detection
Curvature \(\rho\) | NEU-seg | MT-defect | FSSD-12 | UCF-EL |
|---|---|---|---|---|
0 | 85.16 | 79.32 | 75.26 | 57.87 |
0.001 | 85.92 | 80.12 | 75.87 | 58.13 |
0.005 | 86.43 | 80.93 | 76.42 | 58.62 |
0.01 | 87.00 | 81.45 | 77.04 | 59.41 |
0.05 | 86.87 | 81.29 | 76.88 | 59.03 |
0.1 | 86.90 | 81.35 | 76.92 | 58.11 |
0.3 | 86.82 | 81.34 | 76.85 | 57.96 |
0.5 | 86.91 | 81.39 | 76.93 | 58.14 |
