Human face detection in still image using Multilayer perceptron solution based on Neuroph framework
R. Vápeník, Ondrej Kainz, Peter Fecil'ak, František Jakab · 2016
Artificial neural networks (ANN) are one of the dominant learning techniques used in the field of artificial intelligence and have significant assets as their properties imitate the behavior of neurons in human brain. In this paper is presented the research focused on ANN, specifically Multilayer perceptron (MPL) with the aim of detection of human face in the still image. This system was implemented in the Java programming language and its utilizing the software framework called Neuroph, which is primary used precisely for neural networks. Evaluation part presents complex testing procedures and introduces possibilities to modify and set the individual parameters that enhance the overall detection process. Novelty of this research is in the use of ANN for several reasons, one of these being the utilizing of software framework, testing of the parameters and detailed description and evaluation on what influences the detection process. However the principal strength lies in the utilization of ANN for detection of faces rather than for a recognition. Utilization of Neuroph framework is an alternative to the systems that are focused on the detection of faces in still images and are commonly implemented in different programming languages.