Operationalization of YOLOv11 in ROS 2 for Computer Vision Applied to Human-robot Interaction

Computer vision is widely used in robotics, medicine, and security, with deep neural networks like YOLOv11 excelling in high-accuracy classification and feature extraction. However, operationalizing these algorithms for real-world applications requires structured approaches like MLOps to ensure efficient deployment and management. This work integrates computer vision with robotic systems using ROS 2 and implements MLOps concepts for embedded systems. By combining ROS 2 with YOLOv11, robots can perform complex tasks with high precision, such as patient monitoring, object identification, and medical assistance in healthcare environments. The pose classification model achieved a mAP50 of approximately 60 %, with high accuracy for detecting falls (93 %). For object detection, YOLOv11 achieved high precision for bottle (97 %), can (97 %), and cup (98 %), with an mAP50-95 of 90 %. Real-time processing with ROS 2 confirmed consistent performance and precise object visualization, highlighting the system's suitability for interactive robotic applications. This work demonstrates the effective integration of computer vision and MLOps in robotics, enabling advanced functionalities in real-world environments.