SISTEMA DE RECONOCIMEINTO FACIAL DE GÈNEROS APLICANDO TÈCNICAS MACHINE LEARNING EN ESTABLECIMIENTOS COMERCIALES

Abstract

This paper focuses on how to establish better access control for people entering a shopping center with the objective of identifying existing individuals in a database with records registered within the establishment, using recognition techniques facial applying machine learning, proposing a functional model based on training algorithms combined with the comparative method through an interface with the ability to issue alerts to established contacts as a result of the creation of a prototype that allows performing the required actions. The system was developed and implemented using open source algorithms with libraries that contain embedded Davis pattern facial recognition training with 360 iterations performed, which offer an assertiveness of 99.38%. The implementation of this tool allowed to obtain up to 98% accuracy when performing facial recognition using libraries with established training. The work methodology for the development of the solution is established, as well as the libraries and components used in this work. The system functionalities diagrams are presented and the table of results of the iterations of tests performed is shown. Finally, it is concluded that for the correct validation of the image information to be processed, the facial recognition technique of normalization must be used for image alignment by means of the algorithm of deep learning and training of a CNN convolutional network obtained from the library method DLIB.