Smart Doorlock System with Face Recognition Using YOLOv5 Model

Abstract

A smart doorlock system based on facial recognition using the You Only Look Once (YOLOv5) method, integrated with the Internet of Things (IoT), has been developed to enhance physical access security in specific areas. This system is capable of recognizing individuals' faces in real-time by processing video frames, which are then labeled using the YOLOv5 model. Raspberry Pi serves as the connecting device to integrate the facial recognition system with the smart doorlock, allowing the door to automatically open when a registered face is detected. The system's performance was evaluated using confusion matrix, recall, precision, and F1-score metrics, showing an accuracy rate of 90%. While the results demonstrate satisfactory performance under optimal lighting conditions, challenges related to suboptimal lighting and the need for additional hardware to enhance facial detection performance under certain conditions were identified. Overall, this system holds significant potential for deployment in various biometric security applications, though further development is required to improve its performance in more varied environmental conditions.