AI Powered Proactive Surveillance: Design and Implementation of Weapon Detection using YOLOv8 and Deep Learning Techniques

For real-time, low-false-alarm surveillance, the system integrates YOLOv8 weapon detection with Convolutional Neural Network-based behavioral analysis. It attains 84.1% accuracy for behavior analysis, 87% mAP@50 for weapon identification, and 88% mAP with 32 ms inference time. It gives real-time notifications with location and timestamps by processing CCTV (Closed-circuit television) feeds on edge devices. Because it balances accuracy, speed, and modular integration, it performs better than YOLOv5(You Only Look Once version 5) and YOLOv7(You Only Look Once version) and may be deployed in public areas.