Face Verification on InsightFace Recognition (IFRT) all-to-all 1:1 protocol (test)
90.45Accuracy (African)Glint360K-R100(r=0.1)
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Glint360K-R100(r=0.1)Training Dataset=Glint360K, Backbone=ResNet-100, Sampling rate (r)=0.12020.10 | 90.45 | 94.6 | 93.96 | 63.91 | 88.23 | |
| Glint360K-R100(r=1.0)Training Dataset=Glint360K, Backbone=ResNet-100, Sampling rate (r)=12020.10 | 89.5 | 94.23 | 93.54 | 65.07 | 88.67 | |
| MS1MV3-R134Training Dataset=MS1MV3, Backbone=ResNet-1342020.10 | 81.08 | 89.06 | 87.53 | 38.4 | 74.76 | |
| MS1MV2-R50Training Dataset=MS1MV2, Backbone=ResNet-502020.10 | 71.97 | 83.24 | 79.66 | 22.94 | 56.2 | |
| VGG2-R50Training Dataset=VGG2, Backbone=ResNet-502020.10 | 49.2 | 65.93 | 56.22 | 27.15 | 47.13 | |
| CASIA-R100Training Dataset=CASIA, Backbone=ResNet-1002020.10 | 39.67 | 53.93 | 47.81 | 16.17 | 37.53 |