Performance of the five candidate convolutional neural networks
| NASNetLarge | InceptionResNetV2 | Xception | ResNet152V2 | ResNet101V2 | |
| Loss, validation | 0.526799 | 0.554040 | 0.533278 | 0.793147 | 0.744385 |
| Accuracy, validation | 0.742046 | 0.713636 | 0.730682 | 0.721023 | 0.723295 |
| Loss, test | 0.539906 | 0.557637 | 0.555917 | 0.800661 | 0.726274 |
| Accuracy, test | 0.731818 | 0.721591 | 0.706818 | 0.722727 | 0.718750 |
| Precision, label 0 (moderate) | 0.82 | 0.78 | 0.82 | 0.76 | 0.80 |
| Recall, label 0 (moderate) | 0.76 | 0.81 | 0.71 | 0.85 | 0.76 |
| F1 score, label 0 (moderate) | 0.79 | 0.79 | 0.76 | 0.80 | 0.78 |
| Precision, label 1 (extreme) | 0.60 | 0.61 | 0.56 | 0.63 | 0.58 |
| Recall, label 1 (extreme) | 0.68 | 0.56 | 0.70 | 0.48 | 0.64 |
| F1 score, label 1 (extreme) | 0.64 | 0.58 | 0.62 | 0.54 | 0.61 |
Green colour highlights the best metric, yellow colour highlights the second best metric and red colour highlights the third best metric row-wise. The precision, recall and F1 score are presented as proportions (multiply by 100 to have percentages). The precision, recall and F1 score were computed with the test dataset. Receiver operating characteristic curves for each model are available in online supplemental materials.