Application of Deep Learning for Weapons Detection in Surveillance Videos

Weapon detection is a very serious and intense issue as far as the security and safety of the public in general, no doubt it's a hard and difficult task furthermore, its troublesome when you need to do it automatically or with some of the AI model. Different object detection models are available but in case of weapons detection it is difficult to detect the weapons of distinctive size and shapes along with the different colors of the background. Currently, a great deal of Convolutional Neural Network (CNN) based deep learning approaches are proposed for the recognition and classification in real-time. In this paper, we have done the comparative analysis of the two versions which is a state of the art model called YOLOV3 and YOLOV4 for weapons detection. For training purpose, we create weapons dataset and the images are collected from google images along with a portion of different assets. We annotate the images one by one manually in different formats in light of fact that YOLO needs annotation file in text format and some other models need annotation file in XML format. We trained both the versions on a large data set of weapons and afterward tested their results for comparative analysis. We explained in the paper that YOLOV4 performs obviously superior to the YOLOV3 in terms of processing time and sensitivity yet we can compare these two in precision metric. The implementation details and trained models are made public at this link:https://cutt.ly/5kBEPhM.