List your company
ResearchFormerly WizWand

Facial Expression Recognition on AffectNet

67.49AccuracyPOSTER++
46.180451.712757.24562.7773May 10, 2019Jul 2, 2020Aug 26, 2021Oct 19, 2022Dec 13, 2023Feb 4, 2025Mar 31, 2026
Updated 7mo ago

Evaluation results

MethodLinks
POSTER++FP=Yes2026.0367.49
POSTERFP=Yes2026.0367.31
APViTFP=Yes2026.0366.91
DACLFP=Yes2026.0365.2
EmoDC AMDiTFP=No, cnn=cnc2026.0364.49
Meta-Face2ExpFP=Yes2026.0364.23
EmoDC AMDiTFP=No, cnn=cp2026.0363.86
EfficientFaceFP=Yes2026.0363.7
CPGBackbone=SphereFace-20, Expansion=1x, Reduction=0.003x2019.1063.57
MT-ArcResLoss Function=ArcFace, Training=Re-trained2019.0963
FinetuneBackbone=SphereFace-20, Expansion=4x, Reduction=0x2019.1062.54
MT-ArcVGGLoss Function=ArcFace, Training=Re-trained2019.0962
VTFFFP=Yes2026.0361.85
EAC (Ours)Noise (%)=10, Backbone=ResNet-182022.0761.11
RUL (NeurIPS21)Noise (%)=10, Backbone=ResNet-182022.0760.54
EAC (Ours)Noise (%)=20, Backbone=ResNet-182022.0760.29
EmoDC BaseFP=No2026.0359.63
RAN-ResNet18+Network=ResNet18, Year=2019, Oversampling=true2019.0559.5
RANFP=Yes2026.0359.5
RUL (NeurIPS21)Noise (%)=20, Backbone=ResNet-182022.0759.01
FT-MT-VGGBackbone=MT-VGG, Training=Fine-tuned2019.0959
EAC (Ours)Noise (%)=30, Backbone=ResNet-182022.0758.91
SCN (CVPR20)Noise (%)=10, Backbone=ResNet-182022.0758.6
Weighted-LossNetwork=AlexNet, Year=20182019.0558
AlexNet2019.0958
Train from ScratchBackbone=SphereFace-20, Expansion=4x, Reduction=0x2019.1057.64
SCN (CVPR20)Noise (%)=20, Backbone=ResNet-182022.0757.51
IPA2LT (LTNet)FP=No2026.0357.31
BaselineNoise (%)=10, Backbone=ResNet-182022.0757.24
RUL (NeurIPS21)Noise (%)=30, Backbone=ResNet-182022.0756.93
BaselineNoise (%)=20, Backbone=ResNet-182022.0755.89
[66]Network=ResNet80, Year=2018, RAF-DB added to training=true2019.0555.71
SCN (CVPR20)Noise (%)=30, Backbone=ResNet-182022.0754.6
MT-VGGTraining=Trained on Aff-Wild22019.0954
RAN-ResNet18Network=ResNet18, Year=20192019.0552.97
BaselineNoise (%)=30, Backbone=ResNet-182022.0752.16
Up-SamplingNetwork=AlexNet, Year=20182019.0547