DEVELOPMENT OF AN ARTIFICIAL VISION ALGORITHM FOR THE RECOGNITION OF THE CORRECT USE OF EPP

Abstract

In this study, details the creation of a computer vision algorithm to verify the proper use of personal protective equipment (EPP). Using convolutional neural networks (CNN) together with tools such as Python 3.12 and YOLO V8, the system has been developed to identify in real time whether a person is wearing a mask and a headgear correctly. The accuracy of the system has been high thanks to its thorough training and validation, reaching a 96.57% validation rate. This approach is crucial to ensure that industries comply with Good Manufacturing Practices (GMP), thus guaranteeing consumer health and product quality. The development of the program included the installation and use of libraries such as OpenCV, Roboflow, MTCNN, Matplotlib, Imutils, Numpy and OS, which facilitated the detection of faces and EPPs, with training based on a database of more than 1500 labeled images.