Face Identification on MegaFace (test)
99.27Rank-1 AccuracyCos+UNPG
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Cos+UNPGBackbone=ResNet-1002022.03 | 99.27 | — | — | — | — | — | — | |
| CosFaceSource=Re-implemented, Backbone=ResNet-1002022.03 | 99.08 | — | — | — | — | — | — | |
| Arc+UNPGBackbone=ResNet-1002022.03 | 98.82 | — | — | — | — | — | — | |
| ArcFaceSource=Re-implemented, Backbone=ResNet-1002022.03 | 98.56 | — | — | — | — | — | — | |
| MagFaceSource=Re-implemented, Backbone=ResNet-1002022.03 | 98.51 | — | — | — | — | — | — | |
| Circle-lossSource=Original Paper, Backbone=ResNet-1002022.03 | 98.5 | — | — | — | — | — | — | |
| ArcFaceSource=Original Paper, Backbone=ResNet-1002022.03 | 98.35 | — | — | — | — | — | — | |
| Mag+UNPGBackbone=ResNet-1002022.03 | 98.03 | — | — | — | — | — | — | |
| AdaCosSource=Original Paper, Backbone=Not ResNet-1002022.03 | 97.41 | — | — | — | — | — | — | |
| ArcFaceBackbone=Inception-ResNet [34], Training Data=WebFace [43] + MS1M [9]2019.05 | — | 99.78 | 99.65 | 99.48 | 98.87 | 98.03 | 96.88 | |
| CosFaceBackbone=Inception-ResNet [34], Training Data=WebFace [43] + MS1M [9]2019.05 | — | 99.82 | 99.68 | 99.46 | 98.57 | 97.58 | 95.5 | |
| Dynamic AdaCosBackbone=Inception-ResNet [34], Training Data=WebFace [43] + MS1M [9]2019.05 | — | 99.88 | 99.72 | 99.51 | 99.02 | 98.54 | 97.41 | |
| Fixed AdaCosBackbone=Inception-ResNet [34], Training Data=WebFace [43] + MS1M [9]2019.05 | — | 99.85 | 99.7 | 99.47 | 98.8 | 97.92 | 96.85 | |
| l2-softmaxBackbone=Inception-ResNet [34], Training Data=WebFace [43] + MS1M [9]2019.05 | — | 99.73 | 99.49 | 99.03 | 97.85 | 95.56 | 92.05 |