SaveLives - A Real-Time Threat Detection System
Nishit Anand, Rupesh Kumar Koshariya · 2022
These days CCTV cameras are everywhere - malls, schools, shops, public places etc. But still people are able to commit crimes and run away because the police or the authorities are notified very late and are thus unable to do anything. Guns and knives are the most common weapons used in committing crimes today. We propose SaveLives - a system which can detect guns and knives from video footage in real time and thus notify the authorities instantly, so that they can come at the right moment and intervene. Unlike today's surveillance systems, it doesn't require human supervision. Our system also eliminates false positives by cross checking against registered faces like that of policemen or security guard. After the system detects a gun or a knife, Twilio API is triggered to alert the concerned authorities about a possible terrorist attack or a threat to public safety, so that timely action can be taken. This paper presents a system for detection of guns and knives based on Yolov4. We achieved Mean Average Precision (mAP) score of 90.35% and Average Intersection Over Union (IOU) score of 73.68%. Our proposed framework helps to stop crimes and reduce the damage done by criminals, thus saving lives.