YOLO-SpectraWatch: A Novel Architecture for Rapid Suspect Detection in Crowded Urban Environments

This study introduces the Spectra Watch N euralN et (SWNN), a novel integration of a custom-tuned YOLO (You Only Look Once) object detection model, aimed at advancing the efficiency and accuracy of suspect detection in diverse surveillance environments. Utilizing the urban surveillance video dataset (USVD), comprising over 100,000 annotated images across various urban scenarios, the YoLO-SWNN model is trained and validated. The YOLO-SWNN model demonstrated higher performance, with an accuracy rate of 93%, precision of 92 %, recall of 89%, and an F1-score of 90%. These metrics were achieved through rigorous evaluation across critical parameters such as processing speed, scalability, and environmental adaptability, establishing YOLO-SWNN's superiority over baseline models like Faster R-CNN, SSD MobileNet, and RetinaNet. The study's findings highlight the potential of advanced neural network models in transforming security and surveillance applications, making YOLO-SWNN a pivotal tool for real-time detection in complex environments.