Table 4 The results of CNN, wCA-CNN, MCNN, and wCA-MCNN on the test set of KDEF.
From: Multiscale wavelet attention convolutional network for facial expression recognition
Method | TPs and statistical item | Precision | Recall | F1 score | Accuracy |
|---|---|---|---|---|---|
CNN | 1 | 91.68% | 91.50% | 0.91434 | 91.50% |
2 | 91.21% | 90.82% | 0.90878 | 90.82% | |
3 | 91.69% | 91.50% | 0.91479 | 91.50% | |
4 | 91.87% | 91.84% | 0.91779 | 91.84% | |
5 | 92.73% | 92.52% | 0.92505 | 92.52% | |
Average | 91.836% | 91.636% | 0.916150 | 91.636% | |
MCNN | 1 | 93.15% | 93.20% | 0.93118 | 93.20% |
2 | 93.34% | 92.86% | 0.92858 | 92.86% | |
3 | 94.16% | 93.88% | 0.93790 | 93.88% | |
4 | 93.83% | 93.54% | 0.93551 | 93.54% | |
5 | 93.02% | 92.86% | 0.92804 | 92.86% | |
Average | 93.500% | 93.268% | 0.932242 | 93.268% | |
wCA-CNN | 1 | 92.60% | 92.52% | 0.92480 | 92.52% |
2 | 93.49% | 93.20% | 0.93063 | 93.20% | |
3 | 93.90% | 93.54% | 0.93568 | 93.54% | |
4 | 93.72% | 93.54% | 0.93560 | 93.54% | |
5 | 92.72% | 92.52% | 0.92520 | 92.52% | |
Average | 93.286% | 93.064% | 0.930382 | 93.064% | |
wCA-MCNN | 1 | 94.74% | 94.56% | 0.94519 | 94.56% |
2 | 94.95% | 94.90% | 0.94869 | 94.90% | |
3 | 95.05% | 94.90% | 0.94902 | 94.90% | |
4 | 96.38% | 96.26% | 0.96245 | 96.26% | |
5 | 94.25% | 94.22% | 0.94175 | 94.22% | |
Average | 95.074% | 94.968% | 0.949420 | 94.968% |
