Real-time face detection on a Raspberry PI

Abstract

The article describes the implementation of different face detection algorithms to capture human faces from real-time video frames using a Raspberry PI microprocessor. This article examines this issue, proposes the implementation of two distinct real-time face detection algorithms, and presents a comprehensive architectural design. Used methods include Haar Cascades which is known as Viola-Jones algorithm, and Histogram of Oriented Gradients + Linear Support Vector Machines algorithm. The algorithms are implemented with the help of the OpenCV and Dlib libraries, and the Python programming language was used to build the face detection system. The OpenCV and Dlib libraries include a large number of built-in packages that assist with face detection and conduct face operations separately, resulting in reduced processing time and increased efficiency overall. The results confirm that both methods can detect faces in real time with acceptable accuracy and computation time but there are several differences. The Histogram of Oriented Gradients + Linear Support Vector Machines algorithm.method is much more preferable in terms of accuracy, but the image pyramid construction will be computationally demanding.