MLOps in CPS - A use-case for image recognition in changing industrial settings

Although the application of Machine Learning Operations (MLOps) tools within the Cyber-Physical Systems (CPS) domain has clear advantages, it has rarely been utilized. In this paper, we showcase the usage of MLOps processes and tooling for the purpose of image recognition in dynamic industrial environments. We start by outlining how MLOps can streamline the integration and management of ML models in production, emphasizing the need for automation tools in the evolving IT/OT domains. The core of the paper presents an Industry5.0-motivated use-case involving image recognition, demonstrating the development, testing, and deployment of ML models using MLOps tools. We argue against the challenges of manual processes and show the benefits of automation in enhancing efficiency and reducing error. We present a successful implementation and execution of MLOps for industrial automation, demonstrating significant improvements in operational efficiency in both model training, validation and deployment. © 2024 IEEE.