Smart Industrial Safety using Computer Vision
Rehan Bhana, Haitham Mahmoud, Moad Idrissi
Abstract
More than 2.3 million people worldwide suffer from work-related injuries or illnesses each year, resulting in more than 6,000 deaths per day. Providing an unclear work environment and failing to wear appropriate personal protective equipment have been identified as significant contributors to workplace accidents, making it imperative that employers prioritize workplace safety as a priority. Providing proper personal protective equipment (PPE) and maintaining a well-organized, clearly marked (unsafe) work environment can help prevent inconvenient workplace incidents. Furthermore, it promotes a safe working environment, reduces the likelihood of life-threatening events, and enhances overall business and economic conditions. Therefore, this paper proposes safe, smart manufacturing by implementing computer vision technology to detect appropriate PPE worn by workers and ensure a safe workspace to reduce the risk of human injuries. By utilising computer vision technology, we can identify PPE, such as gloves, helmets, and working forklifts, used by workers in the manufacturing environment. A precision of 80.6% and 86% have been reached using YOLOv8 for all classes in both datasets. In general, an extensive review of both datasets, including five performance metrics, is considered.