Figure 13.

Average testing accuracy of five-fold cross-validation with overall accuracy and accuracy calculated separately for banknote recognition and fitness classification using various CNN architectures with models trained from scratch.