Paper Title
HAND-GUN DETECTION IN ROBBERY ACTIVITIES USING RETINA NET MODEL

Abstract
Abstract - Robbery with a weapon in hand is increasing day by day, and it has grown to be one of the greatest issues facing the world nowadays. There are a variety of detecting devices on the current market, but they are not designed to automatically detect robbery activities based on gun detection and alarm the owner. To address this problem, deep learning-based gun detection for determining the robbery activities in CCTV’s camera footage is used. This work presents a comparative study on gun detection algorithms in terms of backbone architecture (ResNet-50, VGG-19 and VGG-16) in RetinaNet. We have achieved an improved real-time gun detection system with the backbone of ResNet-50 that shows a higher mean average precision (mAP) score compared to the other two backbones. We create our own custom dataset for hand-gun detection. Using the RetinaNet model, the guns can be detected without the use of Sensors and Trip Wires. RetinaNet is primarily made up of the residual network (ResNet), the Feature Pyramid Network (FPN), and Fully Convolutional Networks (FCN). This system can alert the Police Station or Command Center without any human effort in real-time. The experimental results proved that our proposed model with the backbone of ResNet-50 outperforms accurately in detecting guns that achieved in a mAP of 96.52%.. Keywords - Robbery Activities, Retina Net, Res Net, FPN, FCN.