IntelligentGuard: Smart Doorbell with Deep Learning for Secure User Recognition and Instant Notifications

In the modern world, daily activities are heavily reliant on the Internet. This study aims to provide users with a simple, personalized technology that effectively manages visitor interactions. The primary objectives are to operate the doorbell intelligently and notify users about visitors by sending a notification with an image of an unknown visitor. This system introduces a low-cost Internet of Things (IoT) smart doorbell designed to enhance home security, utilizing a Raspberry Pi and a camera sensor. Camera sensor is used to capture images in front of the doorbell, which are then processed by the Raspberry Pi and sent to the server. The pre-trained deep learning model kept on a remote server checks the captured image whether it is known or unknown; if it is unknown, a new photo is taken. The proposed hybrid CNN-LSTM model achieves 98% accuracy on the collected dataset and 96.1% accuracy on the Human Faces dataset from Kaggle. This device is utilized to remotely monitor activities outside the door and receive notifications when visitors approach the doorbell using a user-friendly mobile application. This modern security doorbell requires minimal installation tools and lacks interior wiring. In the proposed system, the ESP32-CAM sensor detects visitors and announces their names if they are recognized as known individuals; otherwise, they are labeled as strangers. All relevant details, including timestamps, are promptly sent to the user's mobile application for real-time monitoring.