HyperAI초신경
홈뉴스최신 연구 논문튜토리얼데이터셋백과사전SOTALLM 모델GPU 랭킹컨퍼런스
전체 검색
소개
HyperAI초신경

Face Verification On Ijb C

평가 지표

TAR @ FAR=1e-3
TAR @ FAR=1e-4
TAR @ FAR=1e-5
model
training dataset

평가 결과

이 벤치마크에서 각 모델의 성능 결과

모델 이름
TAR @ FAR=1e-3
TAR @ FAR=1e-4
TAR @ FAR=1e-5
model
training dataset
Paper TitleRepository
circle loss96.29%93.95%89.60%R100MS1M CleanedCircle Loss: A Unified Perspective of Pair Similarity Optimization
ElasticFace-Cos-96.57%-R100MS1M V2ElasticFace: Elastic Margin Loss for Deep Face Recognition
MagFace++-95.97%90.36%R100MS1MV2MagFace: A Universal Representation for Face Recognition and Quality Assessment
Mag+UNPG--94.7%--Unified Negative Pair Generation toward Well-discriminative Feature Space for Face Recognition
ArcFace--96.07%R100IBUG-500KArcFace: Additive Angular Margin Loss for Deep Face Recognition
FaceNet-----FaceNet: A Unified Embedding for Face Recognition and Clustering
PartialFC-97.97%96.93%R200WebFace42MKilling Two Birds with One Stone:Efficient and Robust Training of Face Recognition CNNs by Partial FC
L2E+IS-sampling97.05%95.49%93.25%MobileFaceNetMS1M V3Rectifying the Data Bias in Knowledge Distillation-
CAFace+AdaFace (WebFace4M)98.0897.3%---Cluster and Aggregate: Face Recognition with Large Probe Set
PFEfuse + match95.49%--SphereFace64MS1M V2Probabilistic Face Embeddings
FFC-97.31%-R100WebFace42MAn Efficient Training Approach for Very Large Scale Face Recognition
ArcFace+CSFM-95.9%94.06%--Controllable and Guided Face Synthesis for Unconstrained Face Recognition
HeadSharing: TH-KD-95.48%93.50%MobileFaceNetMS1M V3It's All in the Head: Representation Knowledge Distillation through Classifier Sharing
MN-vc-----Multicolumn Networks for Face Recognition
AdaFace (WebFace4M)-97.39%---AdaFace: Quality Adaptive Margin for Face Recognition
WebFace42M baseline-97.7%-R100WebFace42MWebFace260M: A Benchmark Unveiling the Power of Million-Scale Deep Face Recognition-
AdaFace (MS1MV3)-97.09%---AdaFace: Quality Adaptive Margin for Face Recognition
CurricularFace -96.1%---CurricularFace: Adaptive Curriculum Learning Loss for Deep Face Recognition
QMagFace97.6296.19%---QMagFace: Simple and Accurate Quality-Aware Face Recognition
Cos+UNPG97.5796.38%94.47%R100MS1MV2Unified Negative Pair Generation toward Well-discriminative Feature Space for Face Recognition
0 of 26 row(s) selected.