Haar Cascade Face Detection

A machine learning object detection method using Haar features and boosted classifiers to identify human faces in real-time video streams.

Paul Viola and Michael Jones pioneered this approach in 2001 (the Viola-Jones algorithm). It calculates Haar-like features (rectangular pixel patterns) via an Integral Image for rapid processing. The training process uses AdaBoost to isolate the most relevant visual markers and arranges them into a cascade of increasingly complex stages. This design allows the system to reject non-face regions almost instantly: simple checks occur first to save CPU cycles. It is the standard for real-time detection on edge devices (like the Raspberry Pi) where power efficiency is a priority over the high-compute demands of modern neural networks.

https://docs.opencv.org/4.x/db/d28/tutorial_cascade_classifier.html