Table 6 The impact of different activation functions on network performance.
From: A new method for Tomicus classification of forest pests based on improved ResNet50 algorithm
Num | ResNet50 | Other Modules | Activation Function | Accuracy | ||
|---|---|---|---|---|---|---|
ReLU | PReLU | Swish | ||||
1 | ✓ | ✓ | ✓ | 87.8% | ||
2 | ✓ | ✓ | ✓ | 92.8% | ||
3 | ✓ | ✓ | ✓ | 89.5% | ||