Deep Learning for Mining Site Safety: A YOLOv5-Based PPE Detection Technology

Traumatic brain injuries caused by falls, electrocution, and collisions with rocks or machinery in the mines are among the primary causes of mine fatalities. As a preventative strategy, the Occupational Safety and Health Administration (OSHA) requires mining companies to enforce and regularly monitor the correct use of Personal Protective Equipment (PPE) in the mines. However, employing mere surveillance cameras to monitor and detect any conduct that breaches personal safety is difficult and mostly inefficient. This project offers a deep learning (DL) model based on the You-Only-Look-Once (YOLOv5) architecture for real-time monitoring of worker PPE compliance from a video stream. This system primarily detects three classes of PPE: helmet, reflector and safety boots. The dataset contains 6,300 images and their respective annotations. The model has been developed and deployed as a web application with a mean average precision (mAP) of 0.85 in identifying the presence of PPE on a person.