A Deep Learning Approach to Door Open and Close Monitoring Systems

Abstract

This paper presents a novel approach to implementing a deep learning neural network-based door open and close monitoring system using an STM32 microcontroller. The system integrates multiple sensors, such as proximity, infrared, and camera inputs, to accurately monitor the door's state in real-time. The neural network architecture includes convolutional layers for image processing, enabling the system to analyze visual input and fully connected layers for handling sensor data processing. The model is trained on image and sensor data to classify the door's status as open or closed. The trained model is optimized using STMicroelectronics' X-CUBE-AI and TensorFlow Lite for Microcontrollers to ensure optimal performance for embedded systems. This allows for efficient, low-power consumption while maintaining high processing accuracy. This system provides reliable, real-time door monitoring and secure control for smart home environments. Experimental results confirm the system's robustness, accuracy, and suitability for low-power embedded applications in a variety of real-world scenarios.