Figure A1.

Comparison of ROC curves for conventional classifiers, DL classifiers, and their respective ensembles obtained by averaging the output probabilities. (a) ROC curves for conventional classifiers applied directly to hand-crafted features, without cascade generalization; (b) ROC curves for the deep learning (DL) models only and their ensemble. The area under the curve (AUC) and the AUC for a false-positive rate higher than 0.4 are both presented for comparison. All values are rounded to three significant digits.