A Human Face Recognition using Alyuda Neurointelligence

Abstract

Nowadays, face recognition has been one of the most popular studies. It is considered as a highly interesting topic to do a study on. With the advancement of today’s technology, face recognition has been used in a wide range of areas. For instance, face recognition is very common in the security industry. The main idea of this study is to identify the best algorithm with the smallest mean squared error (MSE). The analyses were carried out to compare the algorithms with the smallest mean squared error and to improve the previous research on face recognition based on artificial neural networks. The study on face recognition data and their evaluation by neural networks is important in detecting human faces. This study was conducted by using 45 different face images. The architecture for the network was obtained by Alyuda Neurointelligence where the most popular learning algorithm such as Quick Propagation, Conjugate Gradient Descent, Quasi Newton, Limited Memory Quasi Newton, Levenberg-Marquadt, Online Back Propagation and Batch Back Propagation have been implemented and tested to measure the percentage of success. The results indicated that the Adaptive Techniques were extremely useful pattern recognition especially in identifying human faces. The Limited Memory Quasi-Newton becomes the most suitable algorithm to train the human face recognition data with the smallest MSE. Furthermore, this study has shown a strong positive relationship proven by the R-squared and correlation coefficient for all algorithms.