loading page

Performance Evaluation and Comparison of YOLOv4 and Multiple Layers of CNN for Weapon Detection
  • Tahreem Tahir
Tahreem Tahir
COMSATS University Islamabad, COMSATS University Islamabad

Corresponding Author:[email protected]

Author Profile

Abstract

Many people have lost their lives in weapon incidents every year. According to CDC report 124 people died every day in 2020 due to gun violence. Research has been carried out in this field in recent years for object detection using machine learning methodologies. In the following research methodology, we have carried out research on deep learning based automated weapon detection system which allows system to detect weapons such as Knife, Handgun, and rifle automatically. We have implemented object detection model You Only Look Once YOLO V4 and trained it on our datasets. Training showed that You Only Look Once YOLO V4 outperformed convolutional neural network CNN. By using this model for closed circuit televisions CCTV, we can help people in saving their lives and reduce human mass slaughter. Moreover, we proposed a method to improve accuracy of Convolutional Neural Network CNN by adding N times N number of layers. It will reduce the complexity of the model and improve accuracy.