Avaliação da aplicação da inteligência artificial na detecção do uso de Equipamentos de Proteção Individual em ambientes de trabalho

Abstract

Personal Protective Equipment (PPE) provides greater safety at the service of a worker, minimizing the risks of accidents, protecting against diseases related to the work environment, and even preventing fatalities. However, several workers do not use the mandatory PPE, and, consequently, the data related to this situation are alarming. The International Labor Organization (ILO) reported that every 15 seconds, a death is caused due to accidents or diseases arising from the work environment. In addition, 153 workers are involved in labor accidents, accidents that occur due to the exercise of work at the company's service. In addition, 6300 daily deaths are recorded due to accidents or illnesses related to work, totaling around 2.3 million annual deaths (Silva, 2018). To combat this, this work will be directed to the detection of PPE using Deep Learning techniques with the convolutional neural network algorithm for the detection of objects in real time, such as You only look once (YOLO), in which in its fourth scaled version (Scaled-YOLOv4), its inference time (prediction time) reaches up to 12 milliseconds (Wang, 2021). Therefore, this technique will be implemented in the chemical laboratories located at the Federal University of Alagoas.