Table 3.
Models’ performance metrics.
| Models | Accuracy | Precision | Recall | Specificity | F1 Score | AUC |
|---|---|---|---|---|---|---|
| VGG16 | 81.77% | 79.05% | 84.69% | 79.05% | 81.77% | 90.15% |
| ResNet50 | 78.82% | 80.22% | 74.49% | 82.86% | 77.25% | 87.19% |
| Wide ResNet50-2 | 81.77% | 79.61% | 83.67% | 80.00% | 81.59% | 88.07% |
| DenseNet161 | 82.76% | 85.39% | 77.55% | 87.62% | 81.28% | 89.56% |
| DenseNet169 | 80.79% | 81.05% | 78.57% | 82.86% | 79.79% | 89.44% |
| Inception v3 | 80.30% | 79.59% | 79.59% | 80.95% | 79.59% | 88.80% |
| Ensemble (Hard voting) | 81.77% | 82.80% | 78.57% | 84.76% | 80.63% | – |
| Ensemble (Soft voting) | 82.27% | 83.70% | 78.57% | 85.71% | 81.05% | 90.01% |
| Proposed Model | 86.70% | 88.17% | 83.67% | 89.52% | 85.86% | 90.82% |