Table 7 The impact of different improvement modules on the performance of network models.
From: A new method for Tomicus classification of forest pests based on improved ResNet50 algorithm
Num | ResNet50 | PReLU | Improved Residual Block | MobileNetV3Classifier | Accuracy | Params | |
|---|---|---|---|---|---|---|---|
DS | ECA | ||||||
1 | ✓ | 83.3% | 25.6 M | ||||
2 | ✓ | ✓ | ✓ | 85.4% | 1.3 M | ||
3 | ✓ | ✓ | ✓ | 90.2% | 9.8 M | ||
4 | ✓ | ✓ | ✓ | 89.1% | 10.4 M | ||
5 | ✓ | ✓ | ✓ | ✓ | 90.7% | 4.9 M | |
6 | ✓ | ✓ | ✓ | ✓ | 91.7% | 1.4 M | |
7 | ✓ | ✓ | ✓ | ✓ | 91.5% | 10.3 M | |
8 | ✓ | ✓ | ✓ | ✓ | ✓ | 92.8% | 1.6M |
