As part of an honors thesis I have cadets working on using RCNNs for image detection. The easiest place for them to begin is to learn CNNs. In keras there is the mnist data set. It is a data set of handwritten digits. It is a great data set to use to familiarize yourself with what images look like to a computer and how to build CNNs. This post attempts to provide a first time student information on how to use keras to build CNNs.

Logistic regression is the last method I shared with the Army unit. I believe logistic regression is an excellent classification model. You are limited to two outcomes, but it is interpretable.

Linear Discriminate Analysis (LDA) is a great classification method. It is simple and interpretable and can perform as well, if not better, than a lot of other fancier models (neural networks that are not interpretable).