Workplace Safety Monitoring Using CNN for Personal Protective Equipment (PPE) Detection

Abstract

Workplace safety in industrial and construction environments is essential to prevent accidents and protect workers` health. The proper use of Personal Protective Equipment (PPE) is fundamental; however, manual supervision of its use is often inefficient. This study aims to implement a computational algorithm based on Convolutional Neural Networks (CNN) for PPE detection, using computer vision to identify in real time safety equipment such as helmets, vests and boots. A specific dataset was developed with over 2,000 images, and the model was implemented using the Roboflow platform. The best iteration of the network achieved a Mean Average Precision (mAP) of 91.1%, with an accuracy of 91% and a recall of 84.7%. These results highlight the potential of the model to improve the monitoring of compliance with safety standards at work, contributing to the reduction of occupational accidents. The methodology used offers an adaptable tool for monitoring the use of PPE, laying the groundwork for future studies that seek to optimize safety in different industrial sectors.