Abstract:

: Reconnaissance security might be a dreary and tedious cycle. In this, we can assemble a system to computerize the test of concentrating on video reconnaissance. We can look at the video feed in genuine time and see any impossible-to-miss exercises like viciousness or theft. CCTV cameras are situated all around the area for observation and security. We have done a “YOLO V4” object identification adaptation by preparing it on our dataset. The tutoring impacts avow that YOLO V4 outflanks YOLO V3 and regular convolutional neural networks (CNN). Besides, concentrated GPUs or unreasonable calculation sources were not needed in our methodology as we utilized switch reading up for preparing our model. Utilizing this model in our observation device, we can attempt to store human lives and accomplish a decrease in the charge of mass killing. Besides, our proposed gadget additionally might be executed in high-stop reconnaissance and well-being robots to stagger on a weapon or hazardous property to avoid any very assault or danger to human lives.
Keywords— Weapon detection, Surveillance, YOLO, Neural Networks.