Real-Time Rapid Visual Multi-Face Detector

Our project’s primary objective is to implement a technique for visual multi-face detection that not only yields accurate results but also exhibits minimal computational complexity. We utilized Haar features-based Adaboost Cascade Classifier, which combines classical image processing concepts with modern machine learning techniques to enable rapid and accurate multi-face detection. It utilizes the modern technique of “Integral Image Representation” for rapid computation of Haar features, subsequently used by Adaboost for classifier learning. We optimize a cascade of such strong classifiers for faster face detection.

In the next phase of the project we integrated the Python code with a live webcam, enabling real-time face detection with the help of bounding boxes. We also compared its compution time with classical approaches such as Convolutional Neural Networks