Improving Face Recognition for Smart Home Security: A Study of Resnet, Siamese Networks, and Attention Mechanisms

This study presents a smart doorbell system designed to improve home security through advanced deep learning techniques. At its core, the system utilizes the YOLOv5 (You Only Look Once) algorithm for real-time face detection, ensuring accurate and efficient visitor identification. To further strengthen security, the system employs multifactor authentication (MFA), integrating face recognition alongside speech recognition and or/a manual keypad. The study evaluates three neural network architectures for face recognition-ResNet with cosine similarity, a Siamese network with CNN, and an attention-enhanced Siamese network-to determine the most effective model for secure and reliable authentication. Performance comparisons highlight progressive improvements in accuracy and robustness,. Additionally, a convenient interface enables seamless remote monitoring and control, enhancing both security and convenience. By addressing key limitations of existing smart security systems, this research provides insight into improving intelligent home access control.