Face Verification on IJB-B (test)
97.84TAR @ FAR=0.01%ResNet200-TopoFR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ResNet200-TopoFRYear='24, Param (M)=118.8, FLOPs (M)=23.5G, Training Dataset=Glint360K, Latency (ms)=12.242026.04 | 97.84 | — | |
| TransFace-LYear='25, Param (M)=271.6, FLOPs (M)=25.4G, Training Dataset=Glint360K, Latency (ms)=OOM2026.04 | 96.73 | — | |
| PartialFC (0.3)Train Data=WF42M, Backbone=R-100, Venue=CVPR22, Re-implemented status=false2026.06 | 96.47 | — | |
| TransFace-BYear='25, Param (M)=124.5, FLOPs (M)=11.5G, Training Dataset=Glint360K, Latency (ms)=18.202026.04 | 96.46 | — | |
| Q-MarginTrain Data=WF42M, Backbone=R-100, alpha=1.25, s=35, m=0.2, Re-implemented status=true2026.06 | 96.44 | 93.67 | |
| EntMaxTrain Data=WF42M, Backbone=R-100, alpha=1.25, s=64, m=0.5, Re-implemented status=true2026.06 | 96.13 | 93.06 | |
| ArcFaceTrain Data=WF42M, Backbone=R-100, s=64, m=0.5, Re-implemented status=true2026.06 | 96.09 | 92.26 | |
| TransFace-SYear='23, Param (M)=86.7, FLOPs (M)=5.8G, Training Dataset=Glint360K, Latency (ms)=14.312026.04 | 96.05 | — | |
| CosFaceTrain Data=WF42M, Backbone=R-100, s=64, m=0.5, Re-implemented status=true2026.06 | 96.05 | 93.15 | |
| AdaFaceTrain Data=WF4M, Backbone=R-100, Venue=CVPR22, Re-implemented status=false2026.06 | 96.03 | — | |
| QAFaceTrain Data=MS1MV2, Backbone=ResNet†, Venue=WACV23, Re-implemented status=false2026.06 | 95.67 | — | |
| CAFace + AdaFaceTrain Data=WF4M, Backbone=R-100, Venue=NeurIPS22, Re-implemented status=false2026.06 | 95.53 | 92.29 | |
| SparseMaxTrain Data=WF42M, Backbone=R-100, alpha=2, s=1.9, m=0.2, Re-implemented status=true2026.06 | 95.5 | 91.84 | |
| ElasticFaceTrain Data=MS1MV2, Backbone=R-100, Venue=CVPR22, Re-implemented status=false2026.06 | 95.43 | — | |
| ResNet50-ArcFaceYear='22, Param (M)=43.6, FLOPs (M)=6.3G, Training Dataset=Glint360K, Latency (ms)=3.762026.04 | 95.3 | — | |
| FaceLiVTv2-LParam (M)=8.52, FLOPs (M)=309, Training Dataset=Glint360K, Latency (ms)=0.71(17.2× ↓)2026.04 | 95.18 | — | |
| FaceLiVTv2-MParam (M)=7.02, FLOPs (M)=258, Training Dataset=Glint360K, Latency (ms)=0.65(18.8× ↓)2026.04 | 95.02 | — | |
| CurricularFaceTrain Data=MS1MV2, Backbone=R-100, Venue=CVPR20, Re-implemented status=false2026.06 | 94.8 | — | |
| AdafaceModel=R-50-512, Params (M)=49.02025.02 | 94.52 | — | |
| FaceLiVTv2-SParam (M)=4.62, FLOPs (M)=179, Training Dataset=Glint360K, Latency (ms)=0.54(22.7× ↓)2026.04 | 94.51 | — | |
| MagFaceTrain Data=MS1MV2, Backbone=R-100, Venue=CVPR21, Re-implemented status=false2026.06 | 94.51 | 90.36 | |
| Efficient Global Neural Architecture Search (medium)Model=R-34-512, Params (M)=37.12025.02 | 93.22 | — | |
| AdafaceModel=R-18-512, Params (M)=29.42025.02 | 93.11 | — | |
| Efficient Global Neural Architecture Search (small)Model=R-58-256, Params (M)=14.62025.02 | 92.06 | — | |
| FaceLiVTv2-XSParam (M)=2.9, FLOPs (M)=90, Training Dataset=Glint360K, Latency (ms)=0.43(28.5× ↓)2026.04 | 90.67 | — |