Cascading classification is another machine learning algorithm that can be used to train a model from many (hundreds, or even thousands) positive and negative image samples. As we explained earlier, a positive image refers to the image in an object of interest (such as a face, a car, or a traffic signal) that we want our model to learn and later classify or detect. On the other hand, a negative image corresponds to any arbitrary image that does not contain our object of interest. The model trained using this algorithm is referred to as a cascade classifier.

The most important aspect of a cascade classifier, as can be guessed from its name, is its cascading nature of learning and detecting an object using the extracted features. The most widely used features in cascade classifiers, and consequently cascade classifier types, are Haar and local binary pattern (LBP). In this section, we're going to learn how to use existing OpenCV Haar and LBP cascade classifiers to detect faces, eyes, and more in real-time, and then learn how to train our own cascade classifiers to detect any other objects.