Face Recognition on LFW (test)
39Rank-1 PSRCLIP2Protect
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| CLIP2ProtectTarget FR Model=MobileFace2023.06 | 39 | — | — | — | — | 61.2 | |
| TIP-IMTarget FR Model=MobileFace2023.06 | 34 | — | — | — | — | 51.4 | |
| CLIP2ProtectTarget FR Model=FaceNet2023.06 | 27.4 | — | — | — | — | 54 | |
| TIP-IMTarget FR Model=FaceNet2023.06 | 25.2 | — | — | — | — | 56.8 | |
| CLIP2ProtectTarget FR Model=Average2023.06 | 23.4 | — | — | — | — | 51.05 | |
| TI-DIMTarget FR Model=MobileFace2023.06 | 21.6 | — | — | — | — | 39 | |
| TIP-IMTarget FR Model=Average2023.06 | 19.7 | — | — | — | — | 41.9 | |
| TI-DIMTarget FR Model=FaceNet2023.06 | 18 | — | — | — | — | 32.8 | |
| CLIP2ProtectTarget FR Model=IR1522023.06 | 16 | — | — | — | — | 51.2 | |
| TI-DIMTarget FR Model=Average2023.06 | 12.85 | — | — | — | — | 26.25 | |
| TIP-IMTarget FR Model=IR1522023.06 | 11.6 | — | — | — | — | 31.2 | |
| CLIP2ProtectTarget FR Model=IRSE502023.06 | 11.2 | — | — | — | — | 37.8 | |
| MI-FGSMTarget FR Model=FaceNet2023.06 | 9 | — | — | — | — | 18.8 | |
| MI-FGSMTarget FR Model=MobileFace2023.06 | 8.4 | — | — | — | — | 22.4 | |
| TIP-IMTarget FR Model=IRSE502023.06 | 8 | — | — | — | — | 28.2 | |
| TI-DIMTarget FR Model=IR1522023.06 | 7.8 | — | — | — | — | 19.6 | |
| MI-FGSMTarget FR Model=Average2023.06 | 6.15 | — | — | — | — | 16.4 | |
| MI-FGSMTarget FR Model=IRSE502023.06 | 4 | — | — | — | — | 10.2 | |
| TI-DIMTarget FR Model=IRSE502023.06 | 4 | — | — | — | — | 13.6 | |
| MI-FGSMTarget FR Model=IR1522023.06 | 3.2 | — | — | — | — | 14.2 | |
| ArcFaceBackbone=ResNet-50, Training Dataset=cleaned WebFace [43], Input Size=144 x 144, Batch Size=512, Feature Dimension=512, Scaling Parameter (s)=30, Margin Parameter (m)=0.52019.05 | — | 99.55 | 99.37 | 99.43 | 99.45 | — | |
| ArcFaceEvaluation Protocol=Fully supervised2022.11 | — | — | — | — | 99.53 | — | |
| CosFaceBackbone=ResNet-50, Training Dataset=cleaned WebFace [43], Input Size=144 x 144, Batch Size=512, Feature Dimension=512, Scaling Parameter (s)=30, Margin Parameter (m)=0.252019.05 | — | 99.37 | 99.35 | 99.42 | 99.38 | — | |
| Dyna. AdaCosBackbone=ResNet-50, Training Dataset=cleaned WebFace [43], Input Size=144 x 144, Batch Size=512, Feature Dimension=512, Scale parameter type=Dynamic2019.05 | — | 99.73 | 99.72 | 99.68 | 99.71 | — | |
| FaceCycleEvaluation Protocol=Linear evaluation2022.11 | — | — | — | — | 74.12 | — | |
| Fixed AdaCosBackbone=ResNet-50, Training Dataset=cleaned WebFace [43], Input Size=144 x 144, Batch Size=512, Feature Dimension=512, Scale parameter type=Fixed2019.05 | — | 99.63 | 99.62 | 99.55 | 99.6 | — | |
| l2-softmaxBackbone=ResNet-50, Training Dataset=cleaned WebFace [43], Input Size=144 x 144, Batch Size=512, Feature Dimension=512, Scaling Parameter (s)=302019.05 | — | 98.22 | 98.27 | 98.08 | 98.19 | — | |
| LBPEvaluation Protocol=Linear evaluation2022.11 | — | — | — | — | 72.44 | — | |
| MoCoEvaluation Protocol=Linear evaluation2022.11 | — | — | — | — | 65.88 | — | |
| PCLEvaluation Protocol=Linear evaluation2022.11 | — | — | — | — | 79.72 | — | |
| SimCLREvaluation Protocol=Linear evaluation2022.11 | — | — | — | — | 75.97 | — | |
| SoftmaxBackbone=ResNet-50, Training Dataset=cleaned WebFace [43], Input Size=144 x 144, Batch Size=512, Feature Dimension=5122019.05 | — | 93.05 | 92.92 | 93.27 | 93.08 | — | |
| SphereFaceEvaluation Protocol=Fully supervised2022.11 | — | — | — | — | 99.42 | — | |
| VGGEvaluation Protocol=Linear evaluation2022.11 | — | — | — | — | 72.2 | — | |
| VGG-FaceEvaluation Protocol=Fully supervised2022.11 | — | — | — | — | 98.95 | — |