P.P.Eye-Track: Real-Time PPE Detection and Dashboard Analytics for ABC Company Construction Project

Abstract

Construction sites continue to face persistent PPE non-compliance despite regulatory requirements and routine inspections. Manual monitoring and post-incident CCTV reviews remain reactive and difficult to sustain across multiple work zones, highlighting the need for affordable real-time monitoring tools for small and medium-sized firms. This study developed and evaluated P.P.Eye‑TRACK, a YOLOv8-based PPE detection and dashboard analytics system for proactive construction safety monitoring. Using a design and development research approach, the system was trained to detect hard hats, safety vests, and safety shoes from CCTV images, deployed for four weeks at an ABC Company project, and evaluated through validation metrics, field observations, real-time performance testing, usability assessment, cost-benefit analysis, and pilot-final survey comparisons. Results showed strong validation performance, with mAP@0.5:0.95 reaching 85.98%. Field deployment accuracy was moderate at 65.94%, with hard hats detected most reliably and safety shoes most affected by object size, distance, lighting, obstruction, and perspective. Green Zone observations were more accurate than Red Zone, confirming the importance of camera placement. The system met real-time performance thresholds, achieved an excellent SUS score of 90.31, and produced a cost-benefit ratio of 1.55 with a 1.82-year payback period. Findings indicated that P.P.Eye‑TRACK is feasible for proactive PPE compliance monitoring, but implementation should include user training, human verification of alerts, expanded safety-shoe datasets, and continuous refinement under varied conditions. This study contributes to SDG 8 by promoting safer workplaces and supports SDG 9 through affordable AI-driven construction safety management.