I have these figures of accuracy and loss of training and validation sets of custom CNN models. I have some understanding of underfitting/overfitting however I am having trouble applying those concepts here. Any thoughts would be helpful. I’d like to justify my selection and talk about the others in a report.