Paper Title
Weapon Detection in Surveillance Cameras using Deep Learning and Transfer Learning

Abstract
In today’s era of high technology, security is one of the major concerns. Nowadays, security surveillance cameras are stationed everywhere in order to detect any kind of anomalous object. In most of the cases, CCTV footage is used after the occurrence of crime, as post evidence. CCTV operators monitor the scene from CCTV but their attentiveness can easily meander due to long shifts. Effectuality of these cameras can be improved by application of object detection algorithms and image processing. Through this paper, we are presenting a neural-network based weapon detection system. It utilizes the concept of deep learning and transfer learning techniques to deliver a highly efficient approach. In our approach we have applied Transfer Learning with the Inception ResNet network to our own custom dataset, called BVCOE Weapon Classification, to get a stellar accuracy of approximately 94%. This framework will reduce the waste of labor, time, and data resources. This proposition can be utilized as an intelligent surveillance system. Keywords - Deep Learning, Data Augmentation, Fine Tuning, Transfer Learning, Inception ResNet, Weapon Classification