Application of the Convolutional Neural Network Method on Face Recognition for Smart Locker

Abstract

Today, security systems are an important issue for the general public, with theft being the most common crime. The weak security system implemented is no doubt behind the many cases of theft. Biometrics is an opportunity to create a strong security system, because each person has their own unique characteristics, such as fingerprints, voice, irises and facial features. One of the biometrics that is considered strong when building a security system is facial recognition. This research uses a Convolutional Neural Network (CNN) derived from Deep Learning as a facial recognition method to create a facial recognition system for cabinets or warehouses. 8820 data was used in system design, which was divided into training data (80%) and test data (20%), the results of the training process obtained validation accuracy reaching 99.81% and validation loss reaching 0.004 after going through 12 epochs. Then the data training process was carried out using the deep learning method using the CNN (Convolutional Neural Network) model. Then a test analysis is carried out to get the accuracy percentage of the entire system. The tests performed in this study gave the system 87.5% accuracy for identifying one individual in the data set and 100% accuracy for individuals not in the data set (unknown). Testing was also carried out with one person in the material who was not in the material in front of the camera in one frame. As a result, the system can recognize all faces and differentiate between people who are in the dataset and those who are not in the dataset (unknown). For testing with two similar people's faces, it was found that the model created was unable to differentiate between the two faces, where the faces detected by the camera showed the same label, namely aqifa with confidence values ​​of 100% and 99.99% respectively.