Face Recognition on LFW
100AccuracyIR18
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| IR18Attack Type=Control No Pattern, Constraint=None, Pattern ID=None, Generation Model=None2026.03 | 100 | — | — | — | — | |
| IR50Attack Type=Control No Pattern2026.03 | 100 | — | — | — | — | |
| SFAttack Type=Control No Pattern2026.03 | 100 | — | — | — | — | |
| TFAttack Type=Control No Pattern2026.03 | 100 | — | — | — | — | |
| EFAttack Type=Control No Pattern2026.03 | 100 | — | — | — | — | |
| ResNet200-TopoFRYear='24, Param (M)=118.8, FLOPs=23.5G, Training Dataset=Glint360K, Model Scale=Large-Scale, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=12.242026.04 | 99.87 | — | — | — | — | |
| GroupFace2021.11 | 99.85 | — | — | — | — | |
| TransFace-LYear='25, Param (M)=271.6, FLOPs=25.4G, Training Dataset=Glint360K, Model Scale=Large-Scale, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=OOM2026.04 | 99.85 | — | — | — | — | |
| TransFace-BYear='25, Param (M)=124.5, FLOPs=11.5G, Training Dataset=Glint360K, Model Scale=Large-Scale, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=18.202026.04 | 99.85 | — | — | — | — | |
| TransFace-SYear='25, Param (M)=86.7, FLOPs=5.8G, Training Dataset=Glint360K, Model Scale=Large-Scale, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=14.312026.04 | 99.85 | — | — | — | — | |
| VarGFaceNetYear='19, Param (M)=5.0, FLOPs=1.0G, Training Dataset=MS1MV3, Model Scale=Large-Scale, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.832026.04 | 99.85 | — | — | — | — | |
| FC-R50Training Dataset=Glint360k, Backbone=ResNet-502021.05 | 99.83 | — | — | — | — | |
| FC-R100Training Dataset=Webface42M, Backbone=ResNet-1002021.05 | 99.83 | — | — | — | — | |
| F2C-R100Training Dataset=Webface42M, Backbone=ResNet-1002021.05 | 99.83 | — | — | — | — | |
| QMagFaceBackbone=iResNet-1002021.11 | 99.83 | — | — | — | — | |
| ElasticFace-Cos2021.11 | 99.82 | — | — | — | — | |
| DuetFaceVenue=MM '22, PPFR=Yes2024.03 | 99.82 | — | — | — | — | |
| DuetFaceDP/LDP Guarantee=None2026.03 | 99.82 | — | — | — | — | |
| KANFace-0.5Year='25, Param (M)=6.80, FLOPs=397M, Training Dataset=WebFace12M, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=9.982026.04 | 99.82 | — | — | — | — | |
| PFE2021.11 | 99.8 | — | — | — | — | |
| CurricularFace2021.11 | 99.8 | — | — | — | — | |
| ElasticFace-Arc2021.11 | 99.8 | — | — | — | — | |
| EQFace2021.11 | 99.8 | — | — | — | — | |
| PartialFaceVenue=ICCV '23, PPFR=Yes2024.03 | 99.8 | — | — | — | — | |
| PartialFaceDP/LDP Guarantee=None2026.03 | 99.8 | — | — | — | — | |
| TFAttack Type=LFW Baseline2026.03 | 99.8 | — | — | — | — | |
| FaceLiVTv2-LYear=-, Param (M)=8.52, FLOPs=309M, Training Dataset=Glint360K, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.71(14.0× ↓)2026.04 | 99.8 | — | — | — | — | |
| FC-R50Training Dataset=MS1MV2, Backbone=ResNet-502021.05 | 99.78 | — | — | — | — | |
| CosFace2021.11 | 99.78 | — | — | — | — | |
| MinusFaceVenue=ours, PPFR=Yes, Backbone=IR-502024.03 | 99.78 | — | — | — | — | |
| MinusFaceDP/LDP Guarantee=None2026.03 | 99.78 | — | — | — | — | |
| ResNet50-ArcFaceYear='22, Param (M)=43.6, FLOPs=6.3G, Training Dataset=Glint360K, Model Scale=Large-Scale, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=3.762026.04 | 99.78 | — | — | — | — | |
| EdgeFace-SYear='24, Param (M)=3.6, FLOPs=306M, Training Dataset=WebFace12M, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=9.892026.04 | 99.78 | — | — | — | — | |
| FaceLiVTv2-MYear=-, Param (M)=7.02, FLOPs=258M, Training Dataset=Glint360K, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.65(16.7%↓)2026.04 | 99.78 | — | — | — | — | |
| FaceLiVTv2-SYear=-, Param (M)=4.62, FLOPs=179M, Training Dataset=Glint360K, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.54(30.8%↓)2026.04 | 99.78 | — | — | — | — | |
| ArcFace2021.11 | 99.77 | — | — | — | — | |
| ArcFaceVenue=CVPR '19, PPFR=No2024.03 | 99.77 | — | — | — | — | |
| DCTDPVenue=ECCV '22, PPFR=Yes, Privacy Budget=epsilon=12024.03 | 99.77 | — | — | — | — | |
| ArcFaceDP/LDP Guarantee=None2026.03 | 99.77 | — | — | — | — | |
| LDP-SlicingDP/LDP Guarantee=Pixel-level2026.03 | 99.75 | — | — | — | — | |
| SwiftFaceFormer-L3Year='24, Param (M)=28.0, FLOPs=2.0G, Training Dataset=MS1MV3, Model Scale=Large-Scale, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=2.592026.04 | 99.75 | — | — | — | — | |
| KANFace-0.5Year='25, Param (M)=6.80, FLOPs=397M, Training Dataset=Glint360K, Model Scale=Lightweight, Experimental Setup=Retrained from scratch, Latency (ms) on iPhone 15 Pro=9.982026.04 | 99.75 | — | — | — | — | |
| EdgeFace-SYear='24, Param (M)=3.6, FLOPs=306M, Training Dataset=Glint360K, Model Scale=Lightweight, Experimental Setup=Retrained from scratch, Latency (ms) on iPhone 15 Pro=9.892026.04 | 99.75 | — | — | — | — | |
| QMagFaceBackbone=iResNet-502021.11 | 99.73 | — | — | — | — | |
| MagFaceBackbone=iResNet-1002021.11 | 99.73 | — | — | — | — | |
| GhostFaceNetV1-1Year='23, Param (M)=4.1, FLOPs=216M, Training Dataset=MS1MV3, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.782026.04 | 99.73 | — | — | — | — | |
| EdgeFace-XSYear='24, Param (M)=1.77, FLOPs=154M, Training Dataset=WebFace12M, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=5.822026.04 | 99.73 | — | — | — | — | |
| MagFaceBackbone=iResNet-502021.11 | 99.72 | — | — | — | — | |
| F2C-R50Training Dataset=Glint360k, Backbone=ResNet-502021.05 | 99.71 | — | — | — | — | |
| EFAttack Type=LFW Baseline2026.03 | 99.7 | — | — | — | — | |
| FaceLiVTv1-MYear='25, Param (M)=9.8, FLOPs=386M, Training Dataset=Glint360K, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.672026.04 | 99.7 | — | — | — | — | |
| MobileFaceNetYear='21, Param (M)=0.99, FLOPs=440M, Training Dataset=MS1MV2, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.772026.04 | 99.7 | — | — | — | — | |
| FaceLiVTv1-SYear='25, Param (M)=5.89, FLOPs=237M, Training Dataset=Glint360K, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.472026.04 | 99.7 | — | — | — | — | |
| PPFR-FDVenue=AAAI '22, PPFR=Yes2024.03 | 99.69 | — | — | — | — | |
| PPFR-FDDP/LDP Guarantee=None2026.03 | 99.69 | — | — | — | — | |
| SwiftFaceFormer-L1Year='24, Param (M)=11.8, FLOPs=805M, Training Dataset=MS1MV3, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=1.202026.04 | 99.68 | — | — | — | — | |
| MixFaceNets-MYear='21, Param (M)=3.9, FLOPs=626M, Training Dataset=MS1MV2, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.702026.04 | 99.68 | — | — | — | — | |
| GhostFaceNetV2-2Year='23, Param (M)=6.8, FLOPs=77M, Training Dataset=MS1MV3, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.672026.04 | 99.68 | — | — | — | — | |
| GhostFaceNetV1-2Year='23, Param (M)=4.1, FLOPs=60M, Training Dataset=MS1MV3, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.602026.04 | 99.68 | — | — | — | — | |
| SphereFace2021.11 | 99.67 | — | — | — | — | |
| ShuffleFaceNet-1.5Year='21, Param (M)=2.6, FLOPs=577M, Training Dataset=MS1MV2, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.692026.04 | 99.67 | — | — | — | — | |
| PocketNetS256Year='22, Param (M)=0.99, FLOPs=587M, Training Dataset=CASIA-WF, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.882026.04 | 99.66 | — | — | — | — | |
| FaceNetImages=200M, Networks=12018.10 | 99.65 | — | — | — | — | |
| PocketNetM128Year='22, Param (M)=1.68, FLOPs=1.1G, Training Dataset=CASIA-WF, Model Scale=Large-Scale, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.982026.04 | 99.65 | — | — | — | — | |
| KANFace-0.6Year='25, Param (M)=4.74, FLOPs=240M, Training Dataset=WebFace12M, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=6.522026.04 | 99.65 | — | — | — | — | |
| FaceLiVTv2-XSYear=-, Param (M)=2.9, FLOPs=90M, Training Dataset=Glint360K, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.43(35.8%↓)2026.04 | 99.63 | — | — | — | — | |
| MobileFaceNetBackbone=MobileFaceNet, MAdds (M)=221, Params(M)=1.02021.04 | 99.6 | — | — | — | — | |
| SwiftFaceFormer-SYear='24, Param (M)=6.0, FLOPs=485M, Training Dataset=MS1MV3, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.652026.04 | 99.6 | — | — | — | — | |
| SwiftFaceFormer-XSYear='24, Param (M)=3.4, FLOPs=294M, Training Dataset=MS1MV3, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.592026.04 | 99.6 | — | — | — | — | |
| PocketNetM256Year='22, Param (M)=1.75, FLOPs=1.1G, Training Dataset=CASIA-WF, Model Scale=Large-Scale, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.982026.04 | 99.58 | — | — | — | — | |
| PocketNetS128Year='22, Param (M)=0.92, FLOPs=587M, Training Dataset=CASIA-WF, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.882026.04 | 99.58 | — | — | — | — | |
| F2C-R50Training Dataset=MS1MV2, Backbone=ResNet-502021.05 | 99.5 | — | — | — | — | |
| SFAttack Type=LFW Baseline2026.03 | 99.5 | — | — | — | — | |
| DCTDPDP/LDP Guarantee=Block-level2026.03 | 99.48 | — | — | — | — | |
| DeepID2+Images=-, Networks=252018.10 | 99.47 | — | — | — | — | |
| OE-CNNsImages=1.7M, Networks=12018.10 | 99.47 | — | — | — | — | |
| SphereFaceImages=0.5M, Networks=12018.10 | 99.42 | — | — | — | — | |
| SteerFaceVenue=(Ours), Generator=DM, Scale (#IDs × #Imgs)=0.5M (10K × 50)2026.05 | 99.4 | — | — | — | — | |
| CASIAVenue=(Real), Generator=/, Scale (#IDs × #Imgs)=0.49M (10.5K × 47)2026.05 | 99.38 | — | — | — | — | |
| OE-CNNsImages=0.5M, Networks=12018.10 | 99.35 | — | — | — | — | |
| MorphFaceVenue=CVPR 25, Generator=DM, Scale (#IDs × #Imgs)=1.2M (24K × 50)2026.05 | 99.35 | — | — | — | — | |
| SteerFaceVenue=(Ours), Generator=DM, Scale (#IDs × #Imgs)=1.2M (24K × 50)2026.05 | 99.35 | — | — | — | — | |
| IDPerturb*Venue=CVPR 26, Generator=DM, Scale (#IDs × #Imgs)=1.2M (24K × 50)2026.05 | 99.33 | — | — | — | — | |
| QMagFaceBackbone=iResNet-182021.11 | 99.3 | — | — | — | — | |
| FNvAttack Type=Control No Pattern2026.03 | 99.3 | — | — | — | — | |
| SFaceAttack Type=Control No Pattern2026.03 | 99.3 | — | — | — | — | |
| Center LossImages=0.7M, Networks=12018.10 | 99.28 | — | — | — | — | |
| UIFaceVenue=ICLR 25, Generator=DM, Scale (#IDs × #Imgs)=0.5M (10K × 50)2026.05 | 99.27 | — | — | — | — | |
| UIFace*Venue=ICLR 25, Generator=DM, Scale (#IDs × #Imgs)=1.2M (10K × 50)2026.05 | 99.27 | — | — | — | — | |
| MorphFaceVenue=CVPR 25, Generator=DM, Scale (#IDs × #Imgs)=0.5M (10K × 50)2026.05 | 99.25 | — | — | — | — | |
| MagFaceBackbone=iResNet-182021.11 | 99.22 | — | — | — | — | |
| ShuffleFaceNet-0.5Year='21, Param (M)=1.4, FLOPs=67M, Training Dataset=MS1MV2, Model Scale=Lightweight, Experimental Setup=Original, Latency (ms) on iPhone 15 Pro=0.452026.04 | 99.2 | — | — | — | — | |
| LF-CNNsImages=0.7M, Networks=12018.10 | 99.1 | — | — | — | — | |
| AsymmNet-LBackbone=AsymmNet-L, MAdds (M)=41.6, Params(M)=1.12021.04 | 99.1 | — | — | — | — | |
| MBV3-LBackbone=MBV3-L, MAdds (M)=41.2, Params(M)=1.02021.04 | 99.1 | — | — | — | — | |
| IDPerturb*Venue=CVPR 26, Generator=DM, Scale (#IDs × #Imgs)=0.5M (10K × 50)2026.05 | 99.08 | — | — | — | — | |
| FC-MobileTraining Dataset=MS1MV2, Backbone=Mobile2021.05 | 99.04 | — | — | — | — | |
| CemiFaceVenue=NeurIPS 24, Generator=DM, Scale (#IDs × #Imgs)=0.5M (10K × 50)2026.05 | 99.03 | — | — | — | — | |
| IR100Attack Type=Control No Pattern2026.03 | 99 | — | — | — | — | |
| F2C-MobileTraining Dataset=MS1MV2, Backbone=Mobile2021.05 | 98.93 | — | — | — | — |