Table 3 The results of ResNet18, MsC-ResNet18, wCA -ResNet18, and MsC- wCA ResNet18 on the test set of FESR.
From: Multiscale wavelet attention convolutional network for facial expression recognition
Method | TPs and statistical item | Precision | Recall | F1 score | Accuracy |
|---|---|---|---|---|---|
ResNet18 | 1 | 88.32% | 92.92% | 0.90391 | 94.60% |
2 | 90.95% | 91.98% | 0.91348 | 95.62% | |
3 | 91.38% | 92.50% | 0.91928 | 95.47% | |
4 | 91.70% | 91.67% | 0.91671 | 95.18% | |
5 | 92.77% | 90.14% | 0.91385 | 95.47% | |
Average | 91.023% | 91.841% | 0.913446 | 95.268% | |
MsC-ResNet18 | 1 | 92.22% | 89.84% | 0.90943 | 95.91% |
2 | 94.12% | 92.49% | 0.93177 | 96.64% | |
3 | 93.72% | 94.76% | 0.94215 | 96.93% | |
4 | 91.55% | 89.99% | 0.90692 | 95.04% | |
5 | 92.45% | 93.57% | 0.92844 | 96.64% | |
Average | 93.007% | 92.162% | 0.924897 | 96.277% | |
wCA -ResNet18 | 1 | 93.12% | 90.59% | 0.91768 | 95.62% |
2 | 90.85% | 92.98% | 0.91839 | 95.47% | |
3 | 95.00% | 93.87% | 0.94407 | 97.08% | |
4 | 91.16% | 91.19% | 0.91147 | 95.18% | |
5 | 91.03% | 92.16% | 0.91557 | 95.77% | |
Average | 92.615% | 92.266% | 0.923907 | 95.985% | |
MsC- wCA ResNet18 | 1 | 96.17% | 93.85% | 0.94967 | 97.81% |
2 | 96.33% | 95.82% | 0.96070 | 98.10% | |
3 | 95.86% | 94.40% | 0.95062 | 97.08% | |
4 | 95.94% | 95.88% | 0.95868 | 98.10% | |
5 | 95.57% | 93.55% | 0.94492 | 97.23% | |
Average | 95.974% | 94.702% | 0.952918 | 97.664% |
