Abstract

To more effectively address the wide range of safety risks faced by construction workers on job sites, machine learning (ML)–based computer vision and augmented reality (AR) technologies are increasingly being employed to enhance efficiency, safety, and productivity. However, current AR construction safety tools only provide passive information for the user to then decide how to use that information. This study leverages advanced computer vision coupled with AR to work with site managers and on-site workers to make operational safety decisions using real-time, visual information of potential hazards. A YOLOv11 model trained to detect the presence or lack of personal protective equipment was developed and tested by creating a local ML computing environment using camera feeds. The detection results were compiled and displayed in real time on a web-based interface developed with Hypertext Preprocessor and on a Microsoft HoloLens 2 heads-up display. The system was successfully field-tested on a construction site.

IPC Classification

Keywords