Published Paper Details:

Today, most criminal activities take place using handheld arms, particularly guns, pistols and revolvers. Crime and social offence can be reduced by monitoring such activities and identifying antisocial behavior so that law enforcement agencies can take appropriate action in an early stage. The solution to this problem is to deploy surveillance systems or control cameras with automatic handheld gun detection and alert systems. In the last few years, deep learning has established a landmark in the area of machine learning particularly in object detection, classification and image segmentation. Convolutional Neural Network (CNN) achieved the best results so far in classical image processing problems such as image segmentation, classification and detection in several applications. Some other problems that arise during gun detection are noises in gun image, deformation of gun and real time processing requirement. During the transportation and addition of images, noises can occur in that image. Various types of noises can be there like salt and pepper noise, Raleigh noise, Gaussian noise, etc. and these noises can be introduced during varied conditions of transportation and addition of noises. In this we present an automatic handheld gun detection system using deep learning, particularly YOLO. From the experiments we found that the most promising results are obtained by YOLO based model trained on our database. The model shows a high potential even in low quality images and provides satisfactory results.