Closed-circuit television is otherwise known as Active video surveillance has been widely utilized throughout enormous groups of public and private entities on a whole, on focused prevention of violence, burglary etcetera. But over the years just the presence of these technologies hasn’t been enough in preventing present or active violence. So we have proposed different models that can detect firearm and alert the human operator when a firearm is visible in the image. The performance of the firearm detection model was analyzed using multiple convolutional neural networks (CNN) architecture, finding values up to 99% accuracy in a network configuration based on MobileNet, about 97% accuracy for VGG16, and about 95% accuracy for ResNet.