List your company

Dynamic Facial Expression Recognition on DFEW

82.48WARHumanOmni
52.049659.949867.8575.7502Dec 9, 2023May 7, 2024Oct 4, 2024Mar 3, 2025Jul 31, 2025Dec 28, 2025May 27, 2026

Evaluation results

MethodLinks
HumanOmniMLLM Based=true, MLLM Params=7B2025.1282.4874.86———————
Omni-EmotionMLLM Based=true, MLLM Params=7B2025.1278.3568.8———————
FaVChatMLLM Based=true, MLLM Params=7B2025.1277.8868.17———————
Emotion-LLaMAMLLM Based=true, MLLM Params=7B2025.1277.0664.21———————
R1-OmniMLLM Based=true, MLLM Params=7B, re-implementation=true2025.1276.6769.6952.0155.3960.7————
FineCLIPER, CLIP-ViT-L/16, LLaVA, MLLMYear=20242026.0476.21————1———
FineCLIPER, CLIP-ViT-L/16, LLaVA, MLLMYear=2024, Individual Features=true, Features=Face analysis & parsing, Landmarks2026.0576.21————1———
S2DTunable Params (M)=9, Sampling strategy=Two clips uniform along temporal axis2023.1276.0361.82———————
S2DMLLM Based=false, MLLM Params=-2025.1276.0361.82———————
S2DTunable Params (M)=92023.1275.9862.57———————
S4DStatus=Original, Training Protocol=Large-scale Pre-training with Additional Data2026.0475.8566.37————64.0417.9341.64
ERVMLLM Based=true, MLLM Params=7B2025.1275.8168.8862.0665.573.53————
S4DStatus=ARGen, Training Protocol=Large-scale Pre-training with Additional Data2026.0475.4766.78————63.8119.3142.07
S2DTunable Params (M)=9, Sampling strategy=Two clips uniform along temporal axis, Training strategy=Oversampling2023.1274.8165.45———————
MAE-DFERTunable Params (M)=85, Sampling strategy=Two clips uniform along temporal axis2023.1274.4363.41———————
MAE-DFERMLLM Based=false, MLLM Params=-2025.1274.4363.41———————
Align-DFER, CLIP-ViT-L/14, MLLMYear=20242026.0474.2————2———
Align-DFER, CLIP-ViT-L/14, MLLMYear=2024, Individual Features=true, Features=Face analysis & parsing, Landmarks2026.0574.2————2———
MIGRMLLM Based=true, MLLM Params=7B2025.1273.9367.7868.4870.0688.95————
UMBEnet, CLIPYear=20242026.0473.93————3———
UMBEnet, CLIPYear=2024, Individual Features=true, Features=Brain, Face2026.0573.93————3———
LSGT, ResNet-18Year=20242026.0472.34————4———
LSGT, ResNet-18Year=2024, Individual Features=true, Features=Face2026.0572.34————4———
DFER-CLIPStatus=ARGen, Training Protocol=Foundation Model-based Adaptation2026.0472.3261.26————58.8417.2440.54
DFER-CLIPTunable Params (M)=902023.1271.2559.61———————
DFER-CLIPStatus=Original, Training Protocol=Foundation Model-based Adaptation2026.0471.2559.61————56.2511.7237.81
CLIPERTunable Params (M)=882023.1270.8457.56———————
VE_MD HeatmapYear=2025, Individual Features=false, Features=Pred. Face SR2026.0570.73————5———
VE-MD-HeatmapYear=20262026.0470.31————5———
VE-MD-PersonQueryYear=20262026.0469.98————6———
VE_MD DETRYear=2025, Individual Features=false, Features=Pred. Face SR2026.0569.98————6———
M3DFEL2023.1269.2556.1———————
IALTunable Params (M)=192023.1269.2455.71———————
OUS, CLIPYear=20242026.0468.85————7———
OUS, CLIPYear=2024, Individual Features=true, Features=Face2026.0568.85————7———
NR-DFERNet2023.1268.1954.21———————
Res18_TransformerStatus=ARGen, Training Protocol=Training from Scratch2026.0467.656.7————55.5411.7239.48
VGG13_TransformerStatus=ARGen, Training Protocol=Training from Scratch2026.0466.7755.53————53.228.2838.89
STT2023.1266.6554.58———————
Res18_TransformerStatus=Original, Training Protocol=Training from Scratch2026.0466.5955.6————52.825.5236.92
ESTTunable Params (M)=432023.1265.8553.43———————
Former-DFERTunable Params (M)=182023.1265.753.69———————
Former-DFERMLLM Based=false, MLLM Params=-2025.1265.753.69———————
VGG13_TransformerStatus=Original, Training Protocol=Training from Scratch2026.0465.553.94————49.663.4535.67
CEFLNetTunable Params (M)=132023.1265.3551.14———————
Res18_LSTMStatus=ARGen, Training Protocol=Training from Scratch2026.0465.2253.25————52.86.938.58
ResNet-18+LSTM2023.1263.8551.32———————
Res18_LSTMStatus=Original, Training Protocol=Training from Scratch2026.0463.6651.85————51.02035.91
VGG13_LSTMStatus=ARGen, Training Protocol=Training from Scratch2026.0463.550.7————50.343.4525.97
VGG13_LSTMStatus=Original, Training Protocol=Training from Scratch2026.0462.4749.44————47.62022.1
EmoCLIP, CLIP-ViT-B/32Year=20242026.0462.12————8———
EmoCLIP CLIP-ViT-B/32Year=2024, Individual Features=true, Features=Face2026.0562.12————8———
3D ResNet-18Tunable Params (M)=332023.1258.2746.52———————
C3DTunable Params (M)=782023.1253.5442.74———————
R(2+1)D-18Tunable Params (M)=332023.1253.2242.79———————