Determining the fittest frontal view face AdaBoost classifier for adoption in personage detector

A big challenge in many face detection applications is how well faces can be reliably detected. An increasing number of algorithms have been published and many such face detectors have been made publicly available in the form of open source applications, such as the OpenCV classifiers. In this work, we attempt to compare four frontal face OpenCV classifiers and determine the most fit for adoption in personage detector. We focus on frontal view following the expert´s rule of thumb when looking for any person of interest in an image. We chose AdaBoost classifiers because AdaBoost algorithm often results in improved performance. We measured the detectors´ performance on the accuracy in terms of recall, precision and harmonic mean. We performed testing over four heterogeneous datasets to obtain a reasonably accurate estimation of the performance of the classifiers that will help us in identifying the most suitable classifier meeting our needs.