1:1 Face Verification on IJB-B
96.02TAR (FAR=1e-4)ArcFace
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ArcFaceTraining Data=IBUG-500K, Backbone=ResNet-1002018.01 | 96.02 | — | — | — | |
| ArcFaceTraining Data=MS1MV3, Backbone=ResNet-1002018.01 | 95.42 | — | — | — | |
| Prodpoly-ResNet50Backbone=ResNet-502020.06 | 95.19 | 43.46 | 91.95 | 96.67 | |
| GroupFace2018.01 | 94.93 | — | — | — | |
| CurricularFace2018.01 | 94.86 | — | — | — | |
| ResNet50Backbone=ResNet-502020.06 | 94.73 | 37.28 | 90.73 | 96.63 | |
| BroadFace2018.01 | 94.61 | — | — | — | |
| MagFace+Aggregation Strategy=Weighted by magnitudes2021.03 | 94.51 | 42.32 | 90.36 | — | |
| MagFace2021.03 | 94.33 | 40.91 | 89.88 | — | |
| ArcFace*Source=Quoted from original paper2021.03 | 94.2 | — | — | — | |
| ArcFaceSource=Re-implemented by authors2021.03 | 94.09 | 38.68 | 88.5 | — | |
| CosFace2021.03 | 94.01 | 40.41 | 89.25 | — | |
| MV-Softmax2018.01 | 93.6 | — | — | — | |
| Softmax2021.03 | 90.06 | 46.73 | 75.17 | — | |
| ArcFaceTraining Data=VGG2, Backbone=ResNet-502018.01 | 89.8 | — | — | — | |
| SphereFace2021.03 | 89.19 | 39.4 | 73.58 | — | |
| AFRN2018.01 | 88.5 | — | — | — | |
| DCN2018.01 | 84.9 | — | — | — | |
| SV-AM-Softmax2021.03 | 84.79 | 29.81 | 69.25 | — | |
| MN-vc2018.01 | 83.1 | — | — | — | |
| VGGFace2*Source=Quoted from original paper2021.03 | 80 | — | 67.1 | — | |
| SENet50Backbone=SENet-502018.01 | 80 | — | — | — | |
| ResNet50Backbone=ResNet-502018.01 | 78.4 | — | — | — |