Facial Expression Recognition on RAF (test)
79.4Mean Class RecallCovariance Pooling
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Covariance PoolingRep. Dim.=20002018.07 | 79.4 | — | — | — | — | — | — | — | |
| VGG-FACE (proposed approach)Backbone=VGG-FACE, Training Strategy=Synthesized data augmentation2018.11 | 77.5 | 78.4 | 64.4 | 62.2 | 91.1 | 81.2 | 84.5 | 80.6 | |
| Deep Locality PreservingRep. Dim.=20002018.07 | 74.2 | — | — | — | — | — | — | — | |
| mSVM-DLP-CNNBackbone=DLP-CNN, Classification Protocol=SVM on penultimate features2018.11 | 74.2 | 71.6 | 52.2 | 62.2 | 92.8 | 80.1 | 81.2 | 80.3 | |
| BaselineRep. Dim.=5122018.07 | 71.7 | — | — | — | — | — | — | — | |
| CAKE-3Rep. Dim.=32018.07 | 68.9 | — | — | — | — | — | — | — | |
| Compact ModelRep. Dim.=642018.07 | 67.6 | — | — | — | — | — | — | — | |
| VGG-FACE baselineBackbone=VGG-FACE, Training Strategy=Original training dataset2018.11 | 58.9 | 69.1 | 28.7 | 36.3 | 85.3 | 66.1 | 66.6 | 60 | |
| VGGRep. Dim.=20002018.07 | 58.2 | — | — | — | — | — | — | — | |
| mSVM-VGG-FACEBackbone=VGG-FACE, Classification Protocol=SVM on penultimate features2018.11 | 58.2 | 68.5 | 27.5 | 35.1 | 85.3 | 64.9 | 66.3 | 59.9 | |
| LDA-VGG-FACEBackbone=VGG-FACE, Classification Protocol=LDA on penultimate features2018.11 | 50.6 | 66.1 | 25 | 37.8 | 73.1 | 51.5 | 53.5 | 47.2 |