Face Identification on MegaFace Challenge1 (Identification)
99.13Rank-1 Identification AccuracyPPRN
Evaluation Results
| Method | Links | |
|---|---|---|
| PPRNTraining Dataset=Glint360K, Loss Function=CosFace, Sampling Rate (r)=1.02020.10 | 99.13 | |
| MC-FaceGraphProtocol=Large, Data Refinement=true2018.01 | 99.02 | |
| ArcFaceProtocol=Large, Backbone=ResNet-100, Training Data=IBUG-500K, Data Refinement=true2018.01 | 98.98 | |
| PPRNTraining Dataset=Glint360K, Loss Function=CosFace, Sampling Rate (r)=0.12020.10 | 98.94 | |
| GroupFaceLoss Function=GroupFace2020.10 | 98.74 | |
| GroupFaceProtocol=Large, Data Refinement=true2018.01 | 98.74 | |
| CurricularFaceLoss Function=CurricularFace2020.10 | 98.71 | |
| BroadFaceProtocol=Large, Data Refinement=true2018.01 | 98.7 | |
| DULProtocol=Large, Data Refinement=true2018.01 | 98.6 | |
| ArcFaceProtocol=Large, Backbone=ResNet-100, Training Data=MS1MV3, Data Refinement=true2018.01 | 98.51 | |
| CircleLossLoss Function=CircleLoss2020.10 | 98.5 | |
| SFaceProtocol=Large, Backbone=ResNet-100, Training Dataset=MS1MV2, Data Refinement (R)=true2022.05 | 98.5 | |
| CircleLossProtocol=Large, Data Refinement=true2018.01 | 98.5 | |
| PPRNTraining Dataset=MS1MV2, Loss Function=CosFace, Sampling Rate (r)=1.02020.10 | 98.36 | |
| ArcFaceLoss Function=ArcFace (0.5)2020.10 | 98.35 | |
| ArcFaceProtocol=Large, Backbone=ResNet-100, Training Dataset=MS1MV2, Data Refinement (R)=true2022.05 | 98.35 | |
| PPRNTraining Dataset=MS1MV2, Loss Function=ArcFace, Sampling Rate (r)=1.02020.10 | 98.31 | |
| PPRNTraining Dataset=MS1MV2, Loss Function=ArcFace, Sampling Rate (r)=0.12020.10 | 98.25 | |
| CurricularFaceProtocol=Large, Data Refinement=true2018.01 | 98.25 | |
| PPRNTraining Dataset=MS1MV2, Loss Function=CosFace, Sampling Rate (r)=0.12020.10 | 98.04 | |
| MV-SoftmaxProtocol=Large, Data Refinement=true2018.01 | 98 | |
| CosFaceLoss Function=CosFace (0.35)2020.10 | 97.91 | |
| CosFaceProtocol=Large, Backbone=ResNet-100, Training Dataset=MS1MV2, Data Refinement (R)=true2022.05 | 97.91 | |
| AdaCosProtocol=Large, Data Refinement (R)=true2022.05 | 97.41 | |
| AdocosProtocol=Large, Data Refinement=true2018.01 | 97.41 | |
| P2SGradProtocol=Large, Data Refinement (R)=true2022.05 | 97.25 | |
| P2SGradProtocol=Large, Data Refinement=true2018.01 | 97.25 | |
| Search-SoftmaxProtocol=Large, Data Refinement=true2018.01 | 96.97 | |
| Domain BalancingProtocol=Large, Data Refinement=true2018.01 | 96.35 | |
| SSTProtocol=Large, Data Refinement=true2018.01 | 96.27 | |
| AdaptiveFaceProtocol=Large, Data Refinement (R)=true2022.05 | 95.023 | |
| AdaptiveFaceProtocol=Large, Data Refinement=true2018.01 | 95.02 | |
| ArcFaceProtocol=Small, Backbone=ResNet-50, Training Data=CASIA, Data Refinement=true2018.01 | 91.12 | |
| CosFaceProtocol=Large2022.05 | 82.72 | |
| CosFaceProtocol=Large2018.01 | 82.72 | |
| ArcFaceProtocol=Large, Backbone=ResNet-100, Training Data=IBUG-500K, Data Refinement=false2018.01 | 81.43 | |
| SFaceProtocol=Large, Backbone=ResNet-100, Training Dataset=MS1MV22022.05 | 81.15 | |
| ArcFaceProtocol=Large, Backbone=ResNet-100, Training Dataset=MS1MV22022.05 | 81.03 | |
| ArcFaceProtocol=Large, Backbone=ResNet-100, Training Data=MS1MV3, Data Refinement=false2018.01 | 80.71 | |
| CosFaceProtocol=Large, Backbone=ResNet-100, Training Dataset=MS1MV22022.05 | 80.56 | |
| UniformFaceProtocol=Large2018.01 | 79.98 | |
| PFEProtocol=Large2018.01 | 78.95 | |
| URFaceProtocol=Large2018.01 | 78.6 | |
| Fair LossProtocol=Large2018.01 | 77.45 | |
| ArcFaceProtocol=Small, Backbone=ResNet-50, Training Data=CASIA, Data Refinement=false2018.01 | 77.42 | |
| CosFaceProtocol=Small2018.01 | 77.11 | |
| RegularFaceProtocol=Large2018.01 | 75.61 | |
| SphereFace+Protocol=Small2018.01 | 73.03 | |
| SphereFaceProtocol=Small2018.01 | 72.729 | |
| AM-SoftmaxProtocol=Small2018.01 | 72.47 | |
| FaceNetProtocol=Large2022.05 | 70.49 | |
| FaceNetProtocol=Large2018.01 | 70.49 | |
| RegularFaceProtocol=Small2018.01 | 70.23 | |
| Center LossProtocol=Small2018.01 | 65.49 | |
| Contrastive LossProtocol=Small2018.01 | 65.21 | |
| TripletProtocol=Small2018.01 | 64.79 | |
| SoftmaxProtocol=Small2018.01 | 54.85 |