Development of a Personal Protective Equipment Detection Algorithm Using Computer Vision with Python in State School Construction Sites

The development of this project emerged as a response to the issue of inadequate use of personal protective equipment in Peru, in line with the country’s Law 29783. The primary aim was to create an algorithm for the detection of personal protective equipment through computer vision using Python, specifically on public school construction sites. The chosen algorithm was MASK-RCNN, executed on a Lenovo IdeaPad 5 laptop connected to the rear camera of a Xiaomi 11T smartphone using the DroidCam application to capture images of workers wearing helmets, vests, and closed-toe shoes. The overall accuracy of the algorithm reached 97.22%, calculated by comparing true positive values to the total of 108 values sampled in the confusion matrix, with improvement observed through further training, which involved using images with varying lighting conditions, quality, colors, etc. In summary, the algorithm demonstrated its ability to correctly detect the proper use of the studied personal protective equipment (helmet, vest, and closed-toe shoes).