Table 4.
Comparison of test result for pre-trained and ensemble model for main dataset in one holdout run. Bold indicates the model holds the highest accuracy.
| Model | Accuracy | Precision | Sensitivity | F1 | Specificity |
|---|---|---|---|---|---|
| ResNet50V2 | 0.9347 | 0.9224 | 0.9456 | 0.9339 | 0.9243 |
| DenseNet201 | 0.9347 | 0.9388 | 0.9312 | 0.9350 | 0.9383 |
| MobileNet | 0.9571 | 0.9184 | 0.9956 | 0.9554 | 0.9242 |
| F-EDNC | 0.9755 | 0.9787 | 0.9641 | 0.9713 | 0.9814 |
| O-EDNC | 0.9632 | 0.9551 | 0.9710 | 0.9795 | 0.9630 |
| FC-EDNC | 0.9714 | 0.9837 | 0.9602 | 0.9718 | 0.9833 |
| CANet | 0.9163 | 0.9061 | 0.925 | 0.9155 | 0.908 |