Résumé
As Industry 5.0 continues to evolve, the focus is shifting toward the human center in order to create workplaces that value both productivity and well-being. Achieving this balance depends on collecting ergonomic data that are accurate, and straightforward to analyze. Traditional motion capture systems, based on inertial sensors, provide precise measurements; however, they are expensive, difficult to install, and often uncomfortable for workers. Meanwhile, computer vision has opened up valuable possibilities for markerless motion analysis. However, many available tools are closed-source or difficult to adapt, which limits transparency and reproducibility in research. In this work, we explore a more accessible and open alternative. We substitute the traditional inertial motion capture setup with a new computer vision approach based on YOLOv11, the latest fully open-source model for real-time pose estimation. To validate the effectiveness of the proposed method, a set of movements is performed and recorded simultaneously by inertial sensors and multiple cameras. YOLOv11 is then used to estimate joint angles directly from the videos. The obtained results show that the system based on YOLOv11 and the choice of camera provide motion estimates that are close to those obtained from inertial sensors. This work represents a first exploration of YOLOv11 for ergonomic analysis. By relying on a high-quality and open-source framework, it overcomes many of the limitations of previous proprietary or semi-closed computer vision tools. The proposed approach is practical, affordable, and non-intrusive, making ergonomic evaluation more accessible in real working environments. This is a step forward toward the human-centered and flexible vision of Industry 5.0.