TABLE 6.
The recognition accuracy of pre-network models.
Pre-network model | Input size | Recognition accuracy (%) |
VGG16 | 224 × 224 | 98.48 |
GoogLeNet | 224 × 224 | 98.91 |
ResNet18 | 224 × 224 | 98.92 |
ResNet34 | 224 × 224 | 98.40 |
ResNet50 | 224 × 224 | 97.01 |
ResNet101 | 224 × 224 | 88.61 |
Inception-ResNet-v2 | 224 × 224 | 99.28 |
INSE-ResNet (our work) | 224 × 224 | 99.47 |
The meaning of the “bold” is the best experimental results.