Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{199637,
        author = {Kannika Rani.b.n and T S Prabhakar and Uday R and yashashwini PM and Thejas N},
        title = {Robust Real-Time Weapon Detection and Automated Alert System: Mitigating False Positives Through Dataset Aggregation and YOLOv8 Inference},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {16170-16175},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=199637},
        abstract = {The alarming frequency of armed violence in public and private spaces necessitates an evolution from post-incident video forensics to preemptive, real-time threat-detection systems. While deep learning object detectors have demonstrated considerable promise, they frequently suffer from unacceptably high false-positive rates in complex visual environments, misclassifying benign objects such as closed umbrellas or toy guns as lethal weapons. This paper presents an end-to-end, automated Weapon Detection and Alert System built upon the YOLOv8 architecture, optimized for real-time inference on edge-computing hardware. To minimize false positives, we aggregated three specialized datasets sourced from Kaggle and Rob flow comprising firearms, knives, and deliberately crafted ’look-alike’ confuse objects into a unified training pipeline of approximately 18,000 images. Upon detection, an asynchronous, multi-threaded alert mechanism dispatches annotated forensic snapshots to designated authorities via SMTP (email) and Twilio (SMS) without interrupting the core inference pipeline. We benchmark the proposed system against Efficient-Det-D1, a strong modern baseline. Results confirm that while Efficient Det achieves a higher [email protected] of 91.5%, our aggregated-dataset YOLOv8 model attains superior throughput (52 FPS vs. 28 FPS), a drastically lower false-positive rate of 1.2% compared to 7.1%, and sub-second end-to-end alert latency demonstrating suitability for real-world public safety deployments.},
        keywords = {Weapon Detection, Public Safety, YOLOv8, Efficient Det, Dataset Aggregation, False Positives, Multi-threading, Computer Vision, Real-Time Surveillance, Alert System},
        month = {April},
        }