In the modern world, ensuring public safety in densely populated environments is an increasingly critical challenge due to the rising incidents of violence, theft, and other criminal activities. Traditional surveillance systems, which depend on human operators monitoring multiple video streams, are prone to error, delayed responses, and limited scalability. This study presents the design and development of a real-time automated surveillance system utilizing advanced deep learning technologies. The system integrates YOLOv8 (You Only Look Once) for high-speed, accurate object detection and face recognition algorithms to identify individuals involved in suspicious or violent behavior. The methodology involves capturing live video feeds, preprocessing the data through frame extraction and facial landmark detection, and then applying fine-tuned deep learning models for violence recognition and face identification. The system is equipped with a Streamlit-based dashboard for real-time visualization and incorporates sound and email alert mechanisms to notify security personnel instantly. The model has demonstrated efficient performance in processing live surveillance data with high accuracy and low latency, making it suitable for deployment in crowded public areas such as metro stations, stadiums, and airports. The outcome of this research aims to significantly enhance situational awareness and reduce the response time to critical incidents, thereby promoting a safer public environment through intelligent automation.
Fatima et al. (2025) studied this question.