Table 2.
Comparison of pre-trained DL models and ensemble approach using averaged Precision, Recall, F1-score, and Accuracy over 5 different folds
| Methods | P (%) | R (%) | F(%) | A (%) |
|---|---|---|---|---|
| VGG-16 | 80.18 | 79.17 | 79.01 | 82.22 |
| VGG-19 | 81.84 | 81.90 | 81.03 | 82.94 |
| ResNet-50 | 82.81 | 82.94 | 82.82 | 84.87 |
| ResNet-101 | 82.69 | 81.88 | 82.02 | 84.98 |
| IncepResNetv2 | 83.90 | 83.44 | 83.62 | 85.43 |
| MobileNetV2 | 82.85 | 81.17 | 80.98 | 84.87 |
| InceptionV3 | 82.51 | 82.30 | 82.16 | 84.53 |
| Xception | 85.01 | 85.14 | 85.02 | 86.51 |
| EfficientNet-B0 | 81.60 | 81.34 | 81.40 | 83.96 |
| EfficientNet-B1 | 83.69 | 84.03 | 83.61 | 85.09 |
| EfficientNet-B2 | 82.06 | 82.67 | 82.07 | 83.51 |
| DenseNet-121 | 83.12 | 83.00 | 82.25 | 84.24 |
| DenseNet-169 | 84.07 | 83.74 | 83.83 | 86.06 |
| Ensemble approach | 85.44 | 85.47 | 85.40 | 87.13 |
Bold values denotes the highest performance