Deep Learning Based Unusual Activity Detection and Altering System Using Embedded AI

This project presents a deep learning–based unusual activity detection and alerting system using embedded AI, aimed at improving the efficiency, accuracy, and responsiveness of modern surveillance systems. Traditional video surveillance relies heavily on continuous human monitoring or simple motion-based detection methods, which are often inefficient, error-prone, and unable to understand complex human behaviors. With the rapid increase in surveillance data and the need for faster response to security threats, there is a strong demand for intelligent systems that can automatically analyze video content and identify abnormal activities in real time. The proposed system addresses these challenges by integrating deep learning techniques with embedded hardware to create an automated, reliable, and low-latency monitoring solution. The system operates by capturing live video through a camera installed in the monitoring area. This video stream is processed by an AI processing unit that uses deep learning models to extract both spatial and temporal features from video frames. Convolutional Neural Networks (CNNs) are employed to analyse visual information such as human posture, movement, and object presence, while sequence-based models analyse motion patterns over time. By learning normal behaviour patterns during training, the system is able to detect deviations that indicate unusual or suspicious activities, such as fighting, intrusion, or sudden aggressive movements.