A Handheld Gun Detection using Faster R-CNN Deep Learning
Gyanendra K. Verma, Anamika Dhillon · 2017
Today's, most of the criminal activities are taken place using handheld arms particularly gun, pistol and revolver. Several surveys revealed that hand held gun is the foremost weapon used for diverse crimes like burglary, rape, etc. Therefore, automatic gun detection is a prime requirement in current scenario and this paper presents automatic gun detection from cluttered scene using Convolutional Neural Networks (CNN). We have used Deep Convolutional Network (DCN), a state-of-the-art Faster Region-based CNN model, through transfer learning, for automatic gun detection from cluttered scenes. We have evaluated our gun detection over Internet Movie Firearms Database (IMFDB), a benchmark gun database. For detecting the visual handheld gun, we got propitious performance of our system. Moreover, we demonstrate that, against the number of several training images, CNN model magnifies the classification accuracy, which is most advantageous in those practices where generous liberal is often not available.