In modern society, ensuring public order and safety is of paramount importance. Detecting dangerous weapons like knives and pistols is crucial for preventing violent incidents and ensuring the safety of citizens. With the rise of smart cities, there is a growing need for advanced surveillance systems capable of real-time threat detection. Integrating AI technologies into urban infrastructures can significantly enhance public safety by enabling continuous monitoring and immediate threat identification. This research underscores the significance of utilizing advanced technologies for dangerous weapons detection, focusing on the YOLOv5 architecture. YOLO (You Only Look Once) is a state-of-the-art object detection model known for its real-time processing capabilities, making it highly suitable for applications requiring rapid and accurate detection. This study presents a knowledge distillation technique where YOLOv5x guides YOLOv5s, resulting in a lightweight model, YOLOv5s-x, with improved accuracy. The approach increases mean average precision (mAP) by 3.9%, from 92.3% to 96.2%, while reducing parameters and computations. YOLOv5s-x outperforms other models, like YOLOv8, offering better detection and classification of dangerous weapons for public safety applications.