Facial Expression Recognition on AFEW (test)
60.05AccuracyARPGNet
Evaluation Results
| Method | Links | |
|---|---|---|
| ARPGNetStream type=Multi-stream, Configuration=backbone ablation2025.11 | 60.05 | |
| HSE-NNStream type=Single-stream2025.11 | 59.3 | |
| ARPGNetStream type=Multi-stream, Backbone=default, Fusion method comparison=controlled backbone2025.11 | 57.7 | |
| 5CNNsStream type=Multi-stream2025.11 | 57.4 | |
| MSCMStream type=Single-stream2025.11 | 56.4 | |
| 4CNNs + LMEDStream type=Multi-stream2025.11 | 56.1 | |
| MulTStream type=Multi-stream, Fusion method comparison=controlled backbone2025.11 | 55.87 | |
| Three-level attention with NSTStream type=Single-stream2025.11 | 55.2 | |
| TEMMAStream type=Multi-stream, Fusion method comparison=controlled backbone2025.11 | 54.83 | |
| Two-stageStream type=Single-stream2025.11 | 54.8 | |
| MDANStream type=Single-stream2025.11 | 54.7 | |
| AENStream type=Single-stream2025.11 | 54.6 | |
| Expression Snippet TransformerStream type=Single-stream2025.11 | 54.3 | |
| PACVTStream type=Multi-stream2025.11 | 54 | |
| BLSTMStream type=Single-stream2025.11 | 53.9 | |
| APViTStream type=Single-stream2025.11 | 53.8 | |
| Statistical encodingStream type=Single-stream2025.11 | 53.5 | |
| Multi-attentionStream type=Multi-stream, Fusion method comparison=controlled backbone2025.11 | 53 | |
| DSANStream type=Single-stream2025.11 | 52.7 | |
| PoolingStream type=Single-stream2025.11 | 52.2 | |
| RNN + C3DStream type=Multi-stream2025.11 | 52 | |
| AM-CNNStream type=Multi-stream2025.11 | 51.4 | |
| Graph-TranStream type=Multi-stream2025.11 | 51.2 | |
| Former-DFERStream type=Single-stream2025.11 | 50.9 | |
| URNNStream type=Single-stream2025.11 | 49 | |
| VGG-FACE-LSTM-external-augmentationarchitecture=LSTM, training_data=Union of AFEW and external data, data_augmentation=true2018.11 | 48.6 | |
| VGG-FACE-LSTM-external-augmentationBackbone=VGG-FACE-LSTM, Training Data=AFEW + external data, Augmentation=Yes2018.11 | 48.6 | |
| VGG-FACE (Proposed Approach)use_synthesized_images=true, backbone=VGG-FACE2018.11 | 48.4 | |
| VGG-FACE trained using the proposed approachBackbone=VGG-FACE, Training Method=Facial affect synthesis framework2018.11 | 48.4 | |
| VGG-FACE-FERpre-training=FER2013, fine-tuning=AFEW2018.11 | 48.3 | |
| VGG-FACE-FERBackbone=VGG-FACE, Pre-training=FER20132018.11 | 48.3 | |
| VGG-FACE-externaltraining_data=Union of AFEW and external data2018.11 | 41.4 | |
| VGG-FACE-externalBackbone=VGG-FACE, Training Data=AFEW + external data2018.11 | 41.4 | |
| the VGG-FACE baselinetraining_data=AFEW training set2018.11 | 37.9 | |
| VGG-FACE baselineBackbone=VGG-FACE, Training Data=AFEW training set2018.11 | 37.9 |