Implementasi Sistem Keamanan Pintu Gedung dengan Pendekatan Face Recognition Menggunakan Convolutional Neural Network (CNN) Berbasis IoT

Abstract

In this digital era, technology-based security systems are increasingly needed to increase protection of important buildings and facilities. One technology that is currently developing is facial recognition. Face recognition is a technology used to identify or verify a person's identity through analysis of their facial features. This research aims to improve building security by ensuring only registered people can access the door, reducing the risk of security breaches from physical keys or access cards. This research involves stages of literature study to understand the concepts of face recognition, internet of things and convolutional neural networks, dataset collection, labeling, training, test and evaluation. From the results of the tested data, there are test result values, namely, accuracy value, precision value, recall value, and F1-Score value. These include an accuracy value of 72.7%, a precision value of 50%, a recall value of 75%, and an F1-Score value of 60%. This research shows that facial recognition technology with the Convolutional Neural Network (CNN) algorithm provides maximum facial recognition results in good lighting conditions. However, system performance tends to decrease when lighting is less than optimal.