This chapter covers the less commonly used topics, such as machine learning with multiple classes and GPU-based optimizations. Both the topics are seeing a growth in interest and practical applications, so they deserve a complete chapter. We consider them advanced only as long as additional knowledge is required about machine learning / statistical classification and parallelization. We will start by explaining some of the most well-known classifiers such as KNN, SVM, and Random Forests, all of which are available in the ml module and show how they work with different database formats and multiple classes. Finally, a set of classes and functions to utilize GPU-based computational resources will be described.
Machine learning deals with techniques that allow computers to learn and make decisions by themselves. A central concept in machine learning is the classifier. A classifier learns from the examples in a dataset, where the label of each sample is known. Usually, we have two datasets at hand: training and test. The classifier builds a model using the training set. This trained classifier is expected to predict the label of new unseen samples, so we finally use the test set to validate it and assess label recognition rates.
In this section, we explain the different classes and functions that OpenCV provides for classification, and simple examples of their use. Machine learning classes and functions for statistical classification, regression, and clustering of data are all included in the ml module.