Face Verification on IJB-B
96.9TAR (FAR=1e-4)GIF
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GIFTrain Set=WebFace4M, Backbone=ResNet-1002025.05 | 96.9 | — | — | — | — | — | |
| PFCTraining Dataset=WebFace42M, Backbone=ViT-L, Sampling Ratio=0.32022.03 | 96.71 | — | — | — | — | — | |
| GIFTrain Set=WebFace4M, Backbone=ViT-base2025.05 | 96.68 | — | — | — | — | — | |
| PFCTraining Dataset=WebFace42M, Backbone=ResNet-200, Sampling Ratio=0.32022.03 | 96.57 | — | — | — | — | — | |
| PFCTraining Dataset=WebFace42M, Backbone=ViT-B, Sampling Ratio=0.32022.03 | 96.56 | — | — | — | — | — | |
| TopoFRTraining Data=Glint360K, Backbone=R200, Venue=NeurIPS242024.10 | 96.56 | — | — | — | — | — | |
| PFCTraining Dataset=WebFace42M, Backbone=ResNet-100, Sampling Ratio=0.32022.03 | 96.47 | — | — | — | — | — | |
| TopoFRTraining Data=Glint360K, Backbone=R100, Venue=NeurIPS242024.10 | 96.34 | — | — | — | — | — | |
| PFCTraining Dataset=WebFace12M, Backbone=ResNet-100, Sampling Ratio=0.32022.03 | 96.31 | — | — | — | — | — | |
| AdaFaceTraining Data=Glint360K, Backbone=R200, Venue=CVPR222024.10 | 96.12 | — | — | — | — | — | |
| Part-based FRBackbone=Part fViT-B, Train data=MS1MV32026.06 | 96.11 | — | — | — | — | — | |
| ViT-8RBackbone=ViT-B + 8 Registers, Train data=MS1MV32026.06 | 96.11 | — | — | — | — | — | |
| PPRNTraining Dataset=Glint360K, Loss Function=CosFace, Sampling Rate (r)=1.02020.10 | 96.1 | — | — | — | — | — | |
| PPRNTraining Dataset=Glint360K, Loss Function=CosFace, Sampling Rate (r)=0.12020.10 | 96.1 | — | — | — | — | — | |
| ViTBackbone=ViT-B, Train data=MS1MV32026.06 | 96.05 | — | — | — | — | — | |
| AdaFaceBackbone=ResNet101, Train data=WebFace4M2026.06 | 96.03 | — | — | — | — | — | |
| PFCTraining Dataset=WebFace42M, Backbone=ResNet-100, Sampling Ratio=0.0082022.03 | 96.02 | — | — | — | — | — | |
| TopoFRTraining Data=Glint360K, Backbone=R50, Venue=NeurIPS242024.10 | 95.96 | — | — | — | — | — | |
| ArcFaceTraining Data=Glint360K, Backbone=R200, Venue=CVPR192024.10 | 95.89 | — | — | — | — | — | |
| AdaFaceTraining Data=Glint360K, Backbone=R100, Venue=CVPR222024.10 | 95.87 | — | — | — | — | — | |
| PFCTraining Dataset=WebFace12M, Backbone=ResNet-100, Sampling Ratio=0.0132022.03 | 95.84 | — | — | — | — | — | |
| AdaFaceBackbone=ResNet101, Train data=MS1MV32026.06 | 95.84 | — | — | — | — | — | |
| TopoFRTraining Data=MS1MV2, Backbone=R200, Venue=NeurIPS242024.10 | 95.82 | — | — | — | — | — | |
| TopoFR†Training Data=MS1MV2, Backbone=R200, Venue=NeurIPS242024.10 | 95.77 | — | — | — | — | — | |
| ArcFaceTrain Set=WebFace4M, Backbone=ResNet-1002025.05 | 95.75 | — | — | — | — | — | |
| ArcFaceBackbone=ResNet101, Train data=WebFace4M2026.06 | 95.75 | — | — | — | — | — | |
| AdaFaceTraining Data=MS1MV2, Backbone=R200, Venue=CVPR222024.10 | 95.71 | — | — | — | — | — | |
| TopoFRTraining Data=MS1MV2, Backbone=R100, Venue=NeurIPS242024.10 | 95.7 | — | — | — | — | — | |
| TopoFR†Training Data=MS1MV2, Backbone=R100, Venue=NeurIPS242024.10 | 95.7 | — | — | — | — | — | |
| ArcFaceTraining Data=Glint360K, Backbone=R100, Venue=CVPR192024.10 | 95.69 | — | — | — | — | — | |
| AdaFaceTraining Data=MS1MV2, Backbone=R100, Venue=CVPR222024.10 | 95.67 | — | — | — | — | — | |
| AdaFaceBackbone=ResNet101, Train data=MS1MV22026.06 | 95.67 | — | — | — | — | — | |
| AdaFaceTraining Data=Glint360K, Backbone=R50, Venue=CVPR222024.10 | 95.66 | — | — | — | — | — | |
| ViT-8RBackbone=ViT-B + 8 Registers, Train data=MS1MV22026.06 | 95.66 | — | — | — | — | — | |
| PFCTraining Dataset=WebFace4M, Backbone=ResNet-100, Sampling Ratio=0.32022.03 | 95.64 | — | — | — | — | — | |
| ViT-8RBackbone=ViT-B + 8 Registers, Train data=WebFace4M2026.06 | 95.64 | — | — | — | — | — | |
| AdaFaceTrain Set=WebFace4M, Backbone=ViT-base2025.05 | 95.6 | — | — | — | — | — | |
| VPL2022.03 | 95.56 | — | — | — | — | — | |
| VPL-ArcFaceBackbone=ResNet101, Train data=MS1MV32026.06 | 95.56 | — | — | — | — | — | |
| Virtual FC2022.03 | 95.55 | — | — | — | — | — | |
| ViTBackbone=ViT-B, Train data=MS1MV22026.06 | 95.53 | — | — | — | — | — | |
| WebFace12M2022.03 | 95.5 | — | — | — | — | — | |
| KP-RPEBackbone=ViT-B+KP-RPE, Train data=WebFace4M2026.06 | 95.49 | — | — | — | — | — | |
| TransFace++-BBackbone=ViT-B, Train data=MS1MV22026.06 | 95.45 | — | — | — | — | — | |
| KP-RPE-8R*Backbone=ViT-B+KP-RPE + 8 Registers, Train data=WebFace4M2026.06 | 95.44 | — | — | — | — | — | |
| ElasticFace-Cos+Training Data=MS1MV2, Backbone=R100, Venue=CVPR222024.10 | 95.43 | — | — | — | — | — | |
| ViT-FREE12Exit Depth=12, Fine-tuning Protocol=None, GFLOPs=11.502026.06 | 95.43 | — | — | — | — | — | |
| ElasticFace-Cos+Backbone=ResNet101, Train data=MS1MV22026.06 | 95.43 | — | — | — | — | — | |
| ViT-FREE11Exit Depth=11, Fine-tuning Protocol=None, GFLOPs=10.552026.06 | 95.41 | — | — | — | — | — | |
| ViTBackbone=ViT-B, Train data=WebFace4M2026.06 | 95.31 | — | — | — | — | — | |
| ArcFaceTraining Data=Glint360K, Backbone=R50, Venue=CVPR192024.10 | 95.3 | — | — | — | — | — | |
| ResNet100Bits=FP322022.06 | 95.25 | — | — | — | — | — | |
| Mag+UNPGBackbone=R1002022.03 | 95.21 | 90.93 | 96.5 | — | 46.33 | 97.63 | |
| ArcFaceTraining Data=MS1MV2, Backbone=R200, Venue=CVPR192024.10 | 95.18 | — | — | — | — | — | |
| KP-RPE*Backbone=ViT-B+KP-RPE, Train data=WebFace4M2026.06 | 95.16 | — | — | — | — | — | |
| TopoFRTraining Data=MS1MV2, Backbone=R50, Venue=NeurIPS242024.10 | 95.14 | — | — | — | — | — | |
| ResNet100Quantization data=Real data, Bits=w8a82022.06 | 95.13 | — | — | — | — | — | |
| TopoFR†Training Data=MS1MV2, Backbone=R50, Venue=NeurIPS242024.10 | 95.13 | — | — | — | — | — | |
| DAM-CurricularFaceTraining Data=MS1MV2, Backbone=R100, Venue=ICCV212024.10 | 95.12 | — | — | — | — | — | |
| GIFTrain Set=MS1MV2, Backbone=ResNet-1002025.05 | 95.05 | — | — | — | — | — | |
| Arc+UNPGBackbone=R1002022.03 | 95.04 | 90.57 | 96.6 | — | 44.03 | 97.7 | |
| TransFaceBackbone=TransFace-B, Train data=MS1MV22026.06 | 95.01 | — | — | — | — | — | |
| PPRNTraining Dataset=MS1MV2, Loss Function=CosFace, Sampling Rate (r)=1.02020.10 | 95 | — | — | — | — | — | |
| Cos+UNPGBackbone=R1002022.03 | 94.99 | 90.61 | 96.5 | — | 49.13 | 97.36 | |
| BroadFace2022.03 | 94.97 | — | — | — | — | — | |
| BroadFaceTraining Data=MS1MV2, Backbone=R100, Venue=ECCV202024.10 | 94.97 | — | — | — | — | — | |
| BroadFaceTrain Set=MS1MV2, Backbone=ResNet-1002025.05 | 94.97 | — | — | — | — | — | |
| BroadFaceBackbone=ResNet101, Train data=MS1MV22026.06 | 94.97 | — | — | — | — | — | |
| Sub-center2022.03 | 94.94 | — | — | — | — | — | |
| GroupFaceTrain Set=MS1MV2, Backbone=ResNet-1002025.05 | 94.93 | — | — | — | — | — | |
| PFCTraining Dataset=WebFace4M, Backbone=ResNet-100, Sampling Ratio=0.042022.03 | 94.91 | — | — | — | — | — | |
| GroupFaceLoss Function=GroupFace2020.10 | 94.9 | — | — | — | — | — | |
| ArcFaceBackbone=R1002022.03 | 94.89 | 89.99 | 96.4 | — | 40.68 | 97.59 | |
| CurricularFace2022.03 | 94.83 | — | — | — | — | — | |
| CurricularFaceBackbone=ResNet-100, Training Dataset=MS1MV22020.04 | 94.8 | — | — | — | — | — | |
| CurricularFaceLoss Function=CurricularFace2020.10 | 94.8 | — | — | — | — | — | |
| PPRNTraining Dataset=MS1MV2, Loss Function=ArcFace, Sampling Rate (r)=1.02020.10 | 94.8 | — | — | — | — | — | |
| CosFace2022.03 | 94.8 | — | — | — | — | — | |
| CurricularFaceTraining Data=MS1MV2, Backbone=R100, Venue=CVPR202024.10 | 94.8 | — | — | — | — | — | |
| CosFaceTrain Set=MS1MV2, Backbone=ResNet-1002025.05 | 94.8 | — | — | — | — | — | |
| CurricularFaceTrain Set=MS1MV2, Backbone=ResNet-1002025.05 | 94.8 | — | — | — | — | — | |
| CosFaceBackbone=ResNet101, Train data=MS1MV22026.06 | 94.8 | — | — | — | — | — | |
| CurricularFaceBackbone=ResNet101, Train data=MS1MV22026.06 | 94.8 | — | — | — | — | — | |
| QuantFaceBackbone=ResNet100, Quantization data=Synthetic data, Bits=w8a82022.06 | 94.74 | — | — | — | — | — | |
| SCF-ArcFaceTraining Data=MS1MV2, Backbone=R100, Venue=CVPR212024.10 | 94.74 | — | — | — | — | — | |
| ArcFace+OursTrain Data=cleaned MS1MV2, 3.9M(+70K), Backbone=ResNet-502022.07 | 94.61 | 90.95 | 96.21 | — | — | — | |
| DULTrain Set=MS1MV2, Backbone=ResNet-1002025.05 | 94.61 | — | — | — | — | — | |
| PPRNTraining Dataset=MS1MV2, Loss Function=CosFace, Sampling Rate (r)=0.12020.10 | 94.6 | — | — | — | — | — | |
| MagFaceBackbone=R1002022.03 | 94.51 | 90.36 | — | — | 42.32 | — | |
| MagFace+Training Data=MS1MV2, Backbone=R100, Venue=CVPR212024.10 | 94.51 | — | — | — | — | — | |
| MagFaceTrain Set=MS1MV2, Backbone=ResNet-1002025.05 | 94.51 | — | — | — | — | — | |
| MagFaceBackbone=ResNet101, Train data=MS1MV22026.06 | 94.51 | — | — | — | — | — | |
| AdaFaceTraining Data=MS1MV2, Backbone=R50, Venue=CVPR222024.10 | 94.42 | — | — | — | — | — | |
| PPRNTraining Dataset=MS1MV2, Loss Function=ArcFace, Sampling Rate (r)=0.12020.10 | 94.4 | — | — | — | — | — | |
| CosFaceBackbone=R1002022.03 | 94.39 | 89.38 | 96.17 | — | 42.27 | 97.35 | |
| ArcFace2022.03 | 94.25 | — | — | — | — | — | |
| ArcFaceTrain Set=MS1MV2, Backbone=ResNet-1002025.05 | 94.25 | — | — | — | — | — | |
| ArcFaceBackbone=ResNet101, Train data=MS1MV22026.06 | 94.25 | — | — | — | — | — | |
| ArcFaceBackbone=ResNet-100, Training Dataset=MS1MV22020.04 | 94.2 | — | — | — | — | — | |
| ArcFaceLoss Function=ArcFace (0.5)2020.10 | 94.2 | — | — | — | — | — |