iMagine: AI-Powered Image Analysis for Aquatic Science

The EC-funded iMagine project aims to revolutionise aquatic science research by providing open access to AI-powered image analysis tools and resources. Focused on the theme of "Healthy Oceans, Seas, Coastal and Inland Waters," iMagine addresses the growing need for efficient analysis of vast amounts of image data generated from diverse sources like underwater cameras, drones, microscopes, and satellites.

iMagine operates the iMagine AI Platform", a computational platform built upon the AI4OS framework and supported by AI4EOSC. Hosted on OpenStack clouds within the EGI e-Infrastructure Federation, the platform offers significant GPU and storage capacity to handle the dynamic needs of various research projects. It provides a complete suite of tools for the entire machine learning lifecycle, including image annotation, preprocessing, a deep learning model catalogue, model training and evaluation, and model inference for scientific end-users. This comprehensive approach facilitates collaboration and knowledge sharing between AI experts and aquatic science researchers.

The platform's core functionalities include:

• A generic computational platform supporting the development of AI-based image analysis services for the aquatic science community.
• Development of AI-based image analysis services addressing various scientific challenges within aquatic research.
• Provision of labelled image datasets, enabling AI models' training and retraining.
• Sharing of best practices, disseminating knowledge related to imaging data and AI-driven image analysis in aquatic sciences.

iMagine supports a diverse range of use cases, demonstrating the power of AI for image analysis in aquatic research. These use cases tackle critical issues including floating litter classification and quantification, plankton taxonomic identification, ecosystem statistics generation, oil spill movement and spread prediction, underwater audio data analysis to track vessel activity, and coral reef health monitoring.

The iMagine Competence Centre, consisting of AI experts, domain scientists, and image data owners, facilitates collaboration between use cases and platform providers. The Competence Centre organises regular meetings, training sessions, and feedback collection to refine AI models and ensure the development of robust online services for end-users.

To further enhance data quality, reproducibility, and scientific progress, iMagine adheres to best practices in data management, quality control, and model development. All use cases contribute to publicly available image datasets on Zenodo, allowing for model validation, retraining, and the development of new models. iMagine actively collaborates with other prominent projects like EOSC, AI4EU, and Blue-Cloud 2026 to maximise its impact and promote the broader adoption of AI-powered solutions within the aquatic science community.