Development of a Smart Edge Device for Home Security

Home security remains an essential consideration in this decade, as property crime continues to be a significant concern in Malaysia. Amidst the COVID-19 pandemic, where preventing virus transmission is of utmost importance, integrating home security with the Internet of Things (IoT) technology has become increasingly relevant. This study aims to develop a smart edge device for enhancing the security of houses in residential areas. A real-time facial recognition security system is constructed using You Only Look Once (YOLO) version 7, in conjunction with the Facenet models. These models have been trained, evaluated, and integrated with a Raspberry Pi to establish communication with the user through Node-RED and the Telegram User Interface (UI), utilizing the Message Queuing Telemetry Transport (MQTT) protocol. The performance of the developed facial recognition security system is compared with other state-of-the-art systems in terms of facial recognition accuracy and inference speed on edge devices. The proposed algorithm achieves a remarkable 94.14% accuracy at a speed of 0.8 FPS.