Desarrollo de un Sistema de Visión Artificial para Detectar el Buen y Mal Uso del Arnés de Seguridad en Tiempo Real

In this project, a neuronal network-based computer vision model was created using YOLOv8m to identify the proper and improper use of safety harnesses in industrial environments. Libraries such as OpenCV, NumPy, Pytorch and Ultralytics were used for this purpose, enabling effective model training. The results indicate an accuracy of 91.5% and an F1-score of 88%, demonstrating robust performance. This work supports the automated monitoring of tasks at height with the aim of reducing workplace incidents. Additionally, the future incorporation of the custom YOLOv8m model into real-time monitoring system is proposed, thus reinforcing the safety culture in the workplace.