Year: 2025
Type: article
Abstract: Ensuring workers use personal protective equipment (PPE) correctly is vital for their safety in challenging settings like construction sites, factories, and hospitals. This study presents a system built on YOLOv8, a deep learning technology, designed to identify PPE items like masks, gloves, helmets, and gowns instantly.We trained it with 3,290 labeled images from Roboflow and tested it on a regular laptop (HP 15s with an AMD Ryzen 5 5500U and 8GB RAM) to see how it holds up with basic hardware. When we checked it against a batch of new images (15% of the total), it scored an overall F1 of 89%, doing best with masks at 91% and a bit lower with gloves at 85%. We also tried it out in a workshop, where it caught PPE mistakes in about 2.2 seconds while running smoothly at 30 frames per second. It worked well overall, though it had some trouble in dim light or when people moved fast, especially with spotting gloves. Compared to older methods like Faster R-CNN or SSD, this setup was more accurate and could pick up more types of PPE. The results show that affordable AI tools like this can make a real difference in keeping workplaces safer by automatically checking PPE use.
Authors: Vivek Kumar Gupta, Pratyush Rathore, Puspendra Prajapati, Rajnish Shukla
Institution: Medi-Caps University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4410036173