A Cascade Classifier is a machine-learning method used to quickly detect specific objects or patterns within images or video. It processes an image through a sequence of increasingly complex classification stages, allowing obvious non-matches to be rejected early while only promising regions move to later stages. This makes detection considerably faster than applying a complex analysis to every part of an image.
Cascade classifiers became widely known through the Viola–Jones approach for real-time face detection, using Haar-like features and a cascade of trained classifiers. They can also be applied to detect objects such as eyes or vehicles and are particularly useful in applications where computational resources or real-time performance are important.
