List your company
ResearchFormerly WizWand

Facial Expression Recognition on RAF-DB

93.61AccuracyLaCoVL-FER
70.095676.200382.30588.4097Apr 12, 2021Feb 22, 2022Jan 4, 2023Nov 16, 2023Sep 27, 2024Aug 9, 2025Jun 21, 2026
Updated 4mo ago

Evaluation results

MethodLinks
LaCoVL-FERYear=2026, Vision Backbone=CLIP-ViT-B/162026.0593.61——
FMAEContrastive=false, MIM=true2024.0793.45——
MMPL-FERYear=IF26, Vision Backbone=CLIP-ViT-B/322026.0593.28——
TPRDYear=TAC25, Vision Backbone=CLIP-ViT-B/162026.0592.89——
DCSYear=AAAI25, Vision Backbone=IR50+ViT2026.0592.57——
ICoCOYear=TAC26, Vision Backbone=CLIP-ViT-B/162026.0592.57——
QCSYear=AAAI25, Vision Backbone=IR50+ViT2026.0592.5——
ARBEx2023.0592.47——
CEPromptYear=TCSVT24, Vision Backbone=CLIP-ViT-B/162026.0592.43——
POSTER++2023.0592.21——
POSTER++Year=PR25, Vision Backbone=IR50+ViT2026.0592.21——
VLCAYear=ESWA25, Vision Backbone=CLIP-ViT-B/322026.0592.12——
POSTER2023.0592.05——
POSTERYear=ICCV23, Vision Backbone=IR50+ViT2026.0592.05——
APViTYear=TAC22, Vision Backbone=IR50+ViT-S2026.0591.98——
LA-NetContrastive=true, MIM=false2024.0791.78——
AUNetYear=KBS26, Vision Backbone=ResNet-182026.0591.75——
CLIPERYear=ICME24, Vision Backbone=CLIP-ViT-B/162026.0591.61——
ExpLLMYear=TMM25, Vision Backbone=ViT-L/142026.0591.03——
TransFERYear=ICCV21, Vision Backbone=IR50+ViT-S2026.0590.91——
FG-AGRYear=TCSVT23, Vision Backbone=ResNet-182026.0590.81——
FRAContrastive=true, MIM=false2024.0790.76——
PIDVITYear=TAC22, Vision Backbone=ViT2026.0590.71——
SOFTYear=ECCV22, Vision Backbone=ResNet-182026.0590.42——
CLEFContrastive=true, MIM=false2024.0790.09——
NHGYear=PR26, Vision Backbone=ResNet-182026.0590.09——
Age Guided AttentionMethodology=Age Guided Attention2026.0590.09——
MRANYear=TCSVT23, Vision Backbone=ResNet-182026.0590.03——
EACBackbone=ResNet-182022.0789.99——
EACYear=ECCV22, Vision Backbone=ResNet-182026.0589.99——
FDRL2021.0489.47——
AMGSNYear=PR26, Vision Backbone=ViT-B/162026.0589.34——
CMNet2026.0589.11——
RUL2022.0788.98——
RUL2023.0588.98——
DMUE2022.0788.76——
DMUEBackbone=ResNet-182026.0588.76——
DMUEYear=CVPR21, Vision Backbone=ResNet-182026.0588.76——
Meta-Face2Exp2026.0588.54——
MA-Net2026.0588.4——
EfficientFace2023.0588.36——
EfficientFace2026.0588.36——
EfficientFaceYear=AAAI21, Vision Backbone=ResNet-502026.0588.36——
FaRLContrastive=true, MIM=true2024.0788.31——
Baseline+BPT2026.0588.23——
SCN2026.0588.14——
VTFF2026.0588.14——
VTFFYear=TAC21, Vision Backbone=ResNet18+ViT-B2026.0588.14——
KTN2022.0788.07——
KTN2026.0588.07——
EAC (Ours)Noise (%)=10, Backbone=ResNet-182022.0788.02——
DACL2022.0787.78——
DDL2021.0487.71——
TDF-FER2026.0587.68——
MixAugment + FlippingPre-training=AffectNet, Backbone=ResNet-502022.0587.5477.3—
PCARNet2026.0587.52——
TMKD2026.0387.45——
DENet2026.0587.35——
ViT+SE2026.0587.22——
SCN2022.0787.03——
SCN2023.0587.03——
SCNYear=CVPR20, Vision Backbone=ResNet-182026.0587.03——
SCN2021.0487.01——
SPDNet2021.0487——
Ad-Corre2022.0586.96——
Teacherrole=teacher2026.0386.96——
RKD2026.0386.96——
Baseline2021.0486.93——
RAN2021.0486.9——
RAN2022.0586.9——
RAN2022.0786.9——
RAN2023.0586.9——
RAN2026.0586.9——
RANYear=TIP20, Vision Backbone=ResNet-182026.0586.9——
MCFContrastive=true, MIM=false2024.0786.86——
CERN2026.0586.82——
IPA2LT2021.0486.77——
IPA2LTTraining protocol=trained with both AffectNet and RAF-DB2022.0786.77——
IPA2LT2026.0586.77——
TeKAP2026.0386.31——
RUL (NeurIPS21)Noise (%)=10, Backbone=ResNet-182022.0786.17——
EAC (Ours)Noise (%)=20, Backbone=ResNet-182022.0786.05——
LightExNet2026.0585.97——
KD2026.0385.49——
CATKD2026.0384.68——
SP2026.0384.64——
EAC (Ours)Noise (%)=30, Backbone=ResNet-182022.0784.42——
RUL (NeurIPS21)Noise (%)=20, Backbone=ResNet-182022.0784.32——
DLP-CNN2021.0484.13——
Fully SupervisedNumber of labels=Full dataset (approx. 12k), Backbone=ResNet-18, Pre-training=MS-Celeb-1M2026.0184.13——
SCN (CVPR20)Noise (%)=10, Backbone=ResNet-182022.0782.15——
RUL (NeurIPS21)Noise (%)=30, Backbone=ResNet-182022.0782.06——
Studentrole=student2026.0381.88——
BaselineNoise (%)=10, Backbone=ResNet-182022.0781.01——
SCN (CVPR20)Noise (%)=20, Backbone=ResNet-182022.0779.79——
BaselineNoise (%)=20, Backbone=ResNet-182022.0777.98——
SCN (CVPR20)Noise (%)=30, Backbone=ResNet-182022.0777.45——
BaselineNoise (%)=30, Backbone=ResNet-182022.0775.5——
Single modelModel configuration=single model, Evaluation protocol=transfer learning2026.0671—0.218
EnsembleModel configuration=ensemble, Evaluation protocol=transfer learning2026.06——0.146