Smart Surveillance System for Firearm Detection Using Deep Learning Model

Violence in public places is increasing quickly, and a major issue is that people are carrying dangerous weapons like guns and knives in crowded areas without anyone noticing. Security staff cannot watch every camera feed all the time, which often causes delays in responses and raises the risk to public safety. To address this, we developed a Smart Surveillance System that automatically detects weapons in real time and quickly notifies the authorities so that action can be taken before any harm happens. We used two deep learning models, YOLOv5m and YOLOv8m, training both on the same labelled dataset of 16,000 images that included guns and knives with appropriate bounding box annotations. In testing, YOLOv5m achieved a precision of 80.67%, a recall of 75.72%, and an mAP50 of 81.69%. YOLOv8m performed better with a precision of 87.54%, a recall of 84.08%, and a higher mAP50 of 88.54%. These results show that YOLOv8m is more accurate in detecting weapons, while both models effectively identified guns and knives in real surveillance situations. This system is a practical step toward making public spaces safer by providing automated, continuous, and reliable weapon detection.