List your company

Dynamic Facial Expression Recognition on FERV39k

45.12UARViT-H FlashLite
21.782427.841233.939.9588Dec 9, 2023May 12, 2024Oct 15, 2024Mar 19, 2025Aug 22, 2025Jan 24, 2026Jun 29, 2026

Evaluation results

MethodLinks
ViT-H FlashLitePretrain=U, Modality=V, Face Align.=×, Params (M)=5212026.0645.1255.64———
ViT-L FlashLitePretrain=U, Modality=V, Face Align.=×, Params (M)=3052026.0644.4154.62———
S2DTunable Params (M)=9, Sampling strategy=Two clips uniform along temporal axis, Training strategy=Oversampling2023.1243.9746.21———
ViT-B FlashLitePretrain=U, Modality=V, Face Align.=×, Params (M)=862026.0643.8153.8———
VideoMAE*Pretrain=U, Modality=V, Face Align.=✓, Params (M)=862026.0643.3352.39———
S4DStatus=ARGen, Training Protocol=Large-scale Pre-training with Additional Data2026.0443.3252.1924.5214.2317.12
MAE-DFERTunable Params (M)=85, Sampling strategy=Two clips uniform along temporal axis2023.1243.1252.07———
MAE-DFERPretrain=U, Modality=V, Face Align.=✓, Params (M)=852026.0643.1252.07———
S4DStatus=Original, Training Protocol=Large-scale Pre-training with Additional Data2026.0442.9652.4123.8112.8316.71
DFER-CLIPStatus=ARGen, Training Protocol=Foundation Model-based Adaptation2026.0442.4752.0427.6814.0516.77
SVFAPPretrain=U, Modality=V, Face Align.=✓, Params (M)=782026.0642.1452.29———
TG-DFER2025.1141.551.67———
S2DTunable Params (M)=9, Sampling strategy=Two clips uniform along temporal axis2023.1241.2852.56———
DFER-CLIPTunable Params (M)=902023.1241.2751.65———
DFER-CLIPStatus=Original, Training Protocol=Foundation Model-based Adaptation2026.0441.2751.6525.3911.7814.62
DFER-CLIPPretrain=U, Modality=V+L, Face Align.=✓, Params (M)=1532026.0641.2751.65———
DFER-CLIP2025.1141.2751.65———
CLIPERTunable Params (M)=882023.1241.2351.34———
CLIPERPretrain=U, Modality=V+L, Face Align.=✓2026.0641.2351.34———
CLIPER2025.1141.2351.34———
S2DTunable Params (M)=92023.1240.8951.83———
MIDAS2025.1139.247.37———
ResNet18-ViT2025.1138.3548.43———
LOGO-FormerPretrain=S, Modality=V2026.0638.2248.13———
LOGO-Former2025.1138.2248.13———
AEN2025.1138.1847.88———
VGG13_TransformerStatus=ARGen, Training Protocol=Training from Scratch2026.0438.1147.1723.989.5913.58
STT2023.1237.7648.11———
STTPretrain=S, Modality=V2026.0637.7648.11———
Former-DFERTunable Params (M)=182023.1237.246.85———
Former-DFERPretrain=S, Modality=V, Params (M)=182026.0637.246.85———
Former-DFER2025.1137.246.85———
Res18_TransformerStatus=ARGen, Training Protocol=Training from Scratch2026.0436.946.5223.8214.0612.99
M3DFEL2023.1235.9447.67———
M3DFELPretrain=S, Modality=V2026.0635.9447.67———
M3DFEL2025.1135.9447.67———
VGG13_TransformerStatus=Original, Training Protocol=Training from Scratch2026.0435.8846.0920.857.0711.6
IALTunable Params (M)=192023.1235.8248.54———
IALPretrain=S, Modality=V, Params (M)=192026.0635.8248.54———
IAL2025.1135.8248.54———
Res18_TransformerStatus=Original, Training Protocol=Training from Scratch2026.0435.2945.9622.111.7811.83
Res18_LSTMStatus=ARGen, Training Protocol=Training from Scratch2026.0434.4244.4621.6711.8511.23
EmoCLIPPretrain=U, Modality=V+L2026.0634.2441.46———
VGG13_LSTMStatus=ARGen, Training Protocol=Training from Scratch2026.0434.0145.0718.54.936.26
NR-DFERNet2023.1233.9945.97———
NR-DFERNetPretrain=S, Modality=V2026.0633.9945.97———
NR-DFERNet2025.1133.9945.97———
VGG13_LSTMStatus=Original, Training Protocol=Training from Scratch2026.0432.4443.9517.080.644.18
Res18_LSTMStatus=Original, Training Protocol=Training from Scratch2026.0432.1343.0419.445.789.51
R(2+1)D-18Tunable Params (M)=332023.1231.5541.28———
R(2+1)D182025.1131.5541.28———
ResNet-18+LSTM2023.1230.9242.59———
ResNet18-LSTM2025.1130.9242.95———
I3D-RGB2025.1130.1738.78———
3D ResNet-18Tunable Params (M)=332023.1226.6737.57———
3D ResNet182025.1126.6737.57———
P3D2025.1123.233.39———
C3DTunable Params (M)=782023.1222.6831.69———
C3D2025.1122.6831.69———