Model details | Input image size 224 × 224 |
Data augmentation (used the zoom, width, and shear function) | |
Binary classification with sigmoid activation | |
Adam optimizer (the initial learning rate of 10−5) | |
Batch size 8 | |
Performance | Training accuracy: 92.4% |
Test accuracy: 87.1% | |
Test recall: 86.4% | |
Test precision: 88.9% | |
Test AUC: 0.864 with 95% CI [0.780–0.949] |