Weapon Detection from Surveillance Footage in Real-Time Using Deep Learning
Vedantika Jadhav, Rutuja Deshmukh, Palak Gupta, Sharvi Ghogale, Mahalakshmi Bodireddy · 2023
Weapon violence is a serious threat nowadays due to the rise in technology and criminal intelligence. Manual surveillance is a very tedious task because these activities are atypical in comparison with everyday activities. By introducing machine learning techniques to detect such activities, we can minimise the risk of human errors and prevent details from going unnoticed. Current systems fail when it comes to efficiency of the system and ease of use. We have implemented various models like CNN, YOLOV7, YOLOV8 and VGG for weapon detection and identified their strengths and limitations. A complete system for detection of weapons using these models and raising an alarm has been implemented here to overcome the problems of manual surveillance. Results are presented in a table focusing on metrics like precision and recall as these metrics prove to be more insightful and reliable as compared to accuracy for object detection.