- Home
- Archives
- 2026
- Vol. 7 Issue 3
- Article
SynapTech: A Real-Time Construction Site Safety Monitoring System Using YOLOv8-Based PPE Detection with Automated Alert and Worker Identification
Authors
1 2 3 4 Department of Computer Science and Engineering, PES College of Engineering, Mandya, Karnataka, India. 5 Assistant Professor, Department of Computer Science and Engineering, PES College of Engineering, Mandya, Karnataka, India
Article Information
- Published In
- Volume 7, Issue 3 (2026)
- Published Online
- May-June 2026
- Pages
- 190-198
- Article Type
- Original Article
Abstract
View PDFConstruction sites constitute one of the most hazardous occupational environments globally, with non-compliance with Personal Protective Equipment (PPE) regulations remaining a primary contributor to workplace injuries and fatalities. Manual inspection by safety officers is inherently limited in coverage, scalability, and reliability. This paper presents SynapTech, an end-to-end, camera-based real-time safety monitoring system that leverages the YOLOv8x object detection architecture to automatically verify worker compliance with PPE requirements—including hard hats (blue, red, white, and yellow variants), high-visibility safety vests, face masks, and protective gloves. A two-tier glove detection mechanism combines a dedicated YOLOv8 glove model (glove.pt) with a novel HSV-based skin-tone estimation fallback for hardware-constrained deployments. Upon detecting a non-compliant worker, SynapTech dispatches an asynchronous email alert with an annotated frame capture to the site supervisor within a 10-second cooldown window, preserving full video pipeline throughput. A face-recognition module identifies registered workers by name, and all incidents are logged to a persistent SQLite database for regulatory audit. The trained YOLOv8x model achieves a precision of 0.945 and mAP@50 of 0.929 on the standard Construction Helmet and Vest (CHV) benchmark, outperforming all prior YOLO generations. Training converges across all three loss components with mAP@50 improving from ~0.40 at epoch 1 to ~0.90 at epoch 25. Mean per-frame inference latency of 24.8 ms (~40 FPS) on a Tesla T4 GPU confirms real-time viability. SynapTech operates on standard CCTV infrastructure with no wearable devices, making it a practical, low-cost, and extensible solution for continuous construction site safety enforcement.
Related Articles
Share Article
*Instagram doesn't support direct link sharing from web. Copy the link and share it in your Instagram story or post.