The Dlib library doesn't have many classification algorithms. There are two that are most applicable: KRR and SVM. These methods are implemented as binary classifiers, but for multi-class classification, this library provides the one_vs_one_trainer class, which implements the voting strategy. Note that this class can use classifiers of different types so that you can combine the KRR and the SVM for one classification task. We can also specify which classifiers should be used for which distinct classes.