Face Recognition using Deep Learning for Real-time Intruder Detection

As the incidence of house break-ins continues to rise, the demand for effective real-time intruder detection systems has become increasingly critical. Among various biometric systems, face recognition stands out for its contactless nature, high accuracy, and unique capability to identify individuals. However, current face recognition methods often necessitate substantial computational resources, including high-performance GPUs, which can result in significant costs and complex implementations. This study introduces a resource-efficient face recognition system for real-time intruder detection, leveraging a Raspberry Pi edge device and a face recognition module built upon Dlib’s pretrained deep learning model based on the ResNet architecture. Our methodology comprises two main phases: prototype development and system evaluation. The prototype development phase encompasses hardware configuration, software integration, dataset preparation, individual dataset learning, face matching, and the integration of Twilio API and buzzer systems for real-time responses. Upon detecting an unknown face, the system promptly sends WhatsApp alert messages and activates a buzzer. Comprehensive experime