Face Verification on RFW
98.57Accuracy (Caucasian)false positive rate penalty loss
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| false positive rate penalty lossBackbone=ResNet-100, Training Dataset=BUPT-Globalface2021.06 | 98.57 | 97.4 | 96.48 | 97.37 | 97.45 | 0.85 | |
| false positive rate penalty lossBackbone=ResNet-50, Training Dataset=BUPT-Globalface2021.06 | 98.3 | 96.95 | 96.75 | 96.85 | 97.21 | 0.73 | |
| ArcFaceBackbone=ResNet-100, Training Dataset=BUPT-Globalface2021.06 | 98.17 | 97.32 | 96.1 | 96.68 | 97.07 | 0.89 | |
| ArcFaceBackbone=ResNet-50, Training Dataset=BUPT-Globalface2021.06 | 97.98 | 96.92 | 96.43 | 96.23 | 96.89 | 0.78 | |
| false positive rate penalty lossBackbone=ResNet-34, Training Dataset=BUPT-Globalface2021.06 | 97.92 | 96.7 | 95.85 | 95.77 | 96.56 | 0.75 | |
| false positive rate penalty lossBackbone=ResNet-1002021.06 | 97.6 | 96.82 | 95.65 | 97.03 | 96.78 | 0.82 | |
| ArcFaceBackbone=ResNet-1002021.06 | 97.37 | 96.17 | 94.98 | 96.43 | 96.24 | 0.98 | |
| ArcFaceBackbone=ResNet-34, Training Dataset=BUPT-Globalface2021.06 | 97.37 | 95.86 | 94.55 | 93.87 | 95.37 | 1.53 | |
| false positive rate penalty lossBackbone=ResNet-502021.06 | 97.08 | 96.77 | 95.75 | 96.47 | 96.52 | 0.57 | |
| RL-RBN(arc)Backbone=ResNet-34, Training Dataset=BUPT-Globalface2021.06 | 97.08 | 95.63 | 95.57 | 94.87 | 95.79 | 0.93 | |
| false positive rate penalty lossBackbone=ResNet-342021.06 | 96.78 | 96.38 | 95.17 | 95.95 | 96.07 | 0.69 | |
| ArcFaceBackbone=ResNet-502021.06 | 96.68 | 95.47 | 94.95 | 95.55 | 95.66 | 0.73 | |
| CosFaceBackbone=ResNet-34, Training Dataset=BUPT-Globalface2021.06 | 96.63 | 94.68 | 93.5 | 92.17 | 94.25 | 1.9 | |
| PFEBackbone=ResNet-342021.06 | 96.38 | 94.6 | 94.27 | 95.17 | 95.11 | 0.93 | |
| RL-RBN(arc)Backbone=ResNet-342021.06 | 96.27 | 94.68 | 94.82 | 95 | 95.19 | 0.93 | |
| GACBackbone=ResNet-342021.06 | 96.23 | 95.12 | 94.93 | 94.65 | 95.23 | 0.6 | |
| ArcFaceBackbone=ResNet-342021.06 | 96.18 | 94.67 | 93.72 | 93.98 | 94.64 | 1.11 | |
| RL-RBN(cos)Backbone=ResNet-34, Training Dataset=BUPT-Globalface2021.06 | 96.03 | 95.15 | 94.58 | 94.27 | 95.01 | 0.77 | |
| DebFaceBackbone=ResNet-342021.06 | 95.95 | 94.78 | 94.33 | 93.67 | 94.68 | 0.83 | |
| RL-RBN(cos)Backbone=ResNet-342021.06 | 95.47 | 95.15 | 94.52 | 95.27 | 95.1 | 0.41 | |
| CosFaceBackbone=ResNet-342021.06 | 95.12 | 93.93 | 92.98 | 92.93 | 93.74 | 1.03 | |
| Prodpoly-ResNet50Model=Prodpoly-ResNet502020.06 | 0.997 | 0.993 | 0.9895 | 0.9942 | — | — | |
| ResNet50Model=ResNet502020.06 | 0.9933 | 0.9857 | 0.9833 | 0.9865 | — | — |