Table 7.
Average test result for pre-trained model for the main dataset in five runs. Bold means these three models have the highest accuracy.
| Model | Accuracy | Precision | Sensitivity | F1 | Specificity |
|---|---|---|---|---|---|
| DenseNet201 | 0.9388 | 0.9610 | 0.9231 | 0.9412 | 0.9565 |
| VGG16 | 0.8918 | 0.8612 | 0.8792 | 0.8701 | 0.8692 |
| InceptionV3 | 0.8734 | 0.8490 | 0.8489 | 0.8491 | 0.8560 |
| ResNet50 | 0.7285 | 0.7224 | 0.7314 | 0.7269 | 0.7287 |
| ResNet50V2 | 0.9408 | 0.9306 | 0.9502 | 0.9403 | 0.9320 |
| ResNet152V2 | 0.9224 | 0.8980 | 0.9442 | 0.9205 | 0.9027 |
| Xception | 0.8939 | 0.8410 | 0.9406 | 0.8880 | 0.8571 |
| VGG19 | 0.8776 | 0.8980 | 0.8627 | 0.8800 | 0.8936 |
| ResNet101 | 0.7429 | 0.5673 | 0.8742 | 0.6881 | 0.6798 |
| ResNet101V2 | 0.9306 | 0.9061 | 0.9527 | 0.9289 | 0.8764 |
| NASNet | 0.8980 | 0.8530 | 0.9372 | 0.8931 | 0.8652 |
| MobileNetV2 | 0.9020 | 0.9836 | 0.8456 | 0.9094 | 0.9805 |
| MobileNet | 0.9510 | 0.9383 | 0.9661 | 0.9520 | 0.9407 |
| MobileNetV3Small | 0.5000 | 0 | 0 | 0 | 0.5000 |
| InceptionResNetV2 | 0.9020 | 0.8776 | 0.9227 | 0.8996 | 0.8832 |
| EfficientNetB7 | 0.5000 | 1.0000 | 0.5000 | 0.6667 | 0 |