Launched the FLUID-AI4EOSC project for model and data liquidity

AI and ML workflows require assets that are systematically structured, richly annotated, and technically optimised for automated execution. To meet these challenges and address identified bottlenecks in the European research ecosystem, the FLUID-AI4EOSC project has been officially launched.

The project relies on the participation of 28 institutions from 12 countries, which will build on the AI4EOSC initiative, extending its scientific AI DataLab. FLUID-AI4EOSC introduces the idea of Data and Models Liquidity. This novel concept aims to extend the FAIR principles (i.e. data Findability, Accessibility, Interoperability and Reusability) to ensure that both models and data flow seamlessly across diverse platforms and scientific domains.

In the interest of data and models liquidity, by September 2029, the project will accomplish three primary action pillars:

• Establishing a collaborative Competence Centre on AI/ML, in order to provide coordinated training, expert support, and practical resources for researchers and operators;
• Implementing unified semantic and technical interoperability standards to enable frictionless discovery, combination, and reuse of data and models across scientific domains;
• Delivering an accessible and intuitive Scientific AI DataLab that reduces computational complexity, allowing domain scientists to focus directly on scientific discovery.

The NODC at OGS is involved in one of eight use cases drawn from representative European research infrastructures, ensuring that project solutions are scalable, reproducible, and grounded in real research environments. The use case focuses on seafloor litter images and videos and is led by MARIS, a Dutch company specialising in pan-European marine and oceanographic data management services, with the participation of IFREMER, the French Institute for Ocean Science.

MARIS, OGS and IFREMER will build together a curated and annotated dataset from underwater imagery (e.g., ROV videos, photo mosaics) to develop and train deep-learning models for automated detection and classification of seafloor litter. The data stewards of the NODC will share their consolidated experience in marine litter data management at European and global levels to ensure that the dataset complies with international standards for litter monitoring practices and data management. Furthermore, the working group will develop a prototype litter identification service to facilitate the automatic detection of seafloor litter from underwater imagery.

Maria Eugenia Molina Jack, OGS: “Data stewards are increasingly being asked to participate in projects that aim to facilitate data use by AI-powered services. Before FLUID-AI4EOSC, we worked on another use case for the iMagine project, which led to the release of a marine litter assessment service on the iMagine platform. This environment enables experts and citizen scientists to detect and quantify plastic litter floating on the water surface using drone recordings. The use cases developed in projects such as FLUID-AI4EOSC and iMagine are fundamental, as they provide the initial impetus for wider uptake by European data-sharing initiatives such as EMODnet Chemistry, and potentially more broadly within a global framework.”