A Deep Learning-Based CCTV System for Automatic Smoking Detection in Fire Exit Zones
Sami Sadat, Mohammad Irtiza Hossain, Junaid Ahmed Sifat, Suhail Haque Rafi, Md. Waseq Alauddin Alvi, Md. Khalilur Rhaman
TL;DR
The paper addresses real-time smoking detection in fire-exit CCTV feeds under challenging lighting and occlusion conditions. It benchmarks YOLOv8, YOLOv11, and YOLOv12 and proposes an edge-optimized model tailored for low-latency surveillance, evaluated on multiple edge devices with multithreaded inference. The proposed model achieves recall 78.90% and mAP@50 83.70%, with Jetson Xavier NX latencies of 52–97 ms, surpassing baseline YOLO variants and demonstrating practical applicability for public-safety enforcement. Limitations include dataset scale and binary classification, with future work planned around larger datasets and multi-modal cues to reduce false positives.
Abstract
A deep learning real-time smoking detection system for CCTV surveillance of fire exit areas is proposed due to critical safety requirements. The dataset contains 8,124 images from 20 different scenarios along with 2,708 raw samples demonstrating low-light areas. We evaluated three advanced object detection models: YOLOv8, YOLOv11, and YOLOv12, followed by development of a custom model derived from YOLOv8 with added structures for challenging surveillance contexts. The proposed model outperformed the others, achieving a recall of 78.90 percent and mAP at 50 of 83.70 percent, delivering optimal object detection across varied environments. Performance evaluation on multiple edge devices using multithreaded operations showed the Jetson Xavier NX processed data at 52 to 97 milliseconds per inference, establishing its suitability for time-sensitive operations. This system offers a robust and adaptable platform for monitoring public safety and enabling automatic regulatory compliance.