Table 4.
Fine-tuning information and the number of trainable parameters associated with each model used in our study
| Deep CNN model | Fine-tuned from | Total number of trainable parameters |
|---|---|---|
| VGG16 | 9 | 13,569,793 |
| VGG19 | 11 | 17,699,329 |
| MobileNet | 50 | 2,665,473 |
| MobileNetV2 | 77 | 2,064,769 |
| MobileNetV3Small | 117 | 1,371,849 |
| MobileNetV3Large | 134 | 4,028,273 |
| ResNet50 | 87 | 21,364,225 |
| ResNet50V2 | 95 | 21,352,449 |
| ResNet101 | 172 | 30,640,129 |
| ResNet101V2 | 188 | 30,625,793 |
| ResNet152 | 257 | 39,855,617 |
| ResNet152V2 | 282 | 39,836,673 |
| InceptionV3 | 155 | 16,791,489 |
| Xception | 66 | 14,860,313 |
| InceptionResNetV2 | 390 | 41,922,529 |
| DenseNet121 | 213 | 4,632,897 |
| DenseNet169 | 297 | 8,544,833 |
| DenseNet201 | 353 | 12,741,185 |
| EfficientNetB0 | 118 | 3,700,169 |