SFace: Sigmoid-Constrained Hypersphere Loss for Robust Face Recognition

Benchmark Model Rank Results
face-identification-on-megafaceSFace (MS1MV2, ResNet100, refined)#9Accuracy: 98.50%
face-recognition-on-calfwSFace (MS1MV2, ResNet100)#9Accuracy: 96.07
face-recognition-on-cplfwSFace (MS1MV2, ResNet100)#9Accuracy: 93.28
face-recognition-on-lfwSFace (MS1MV2, ResNet100)#12Accuracy: 99.82
face-recognition-on-cfp-fpSFace (CASIA-WebFace, ResNet50)#13Accuracy: 95.81
face-recognition-on-agedb-30SFace (CASIA-WebFace, ResNet50)#14Accuracy: 95.10
face-verification-on-ijb-cSFace (MS1MV2, ResNet100)#34TAR @ FAR=1e-4: 96.11%