iMagine Impact Report

Impact Report 2025 iMagine Impact Report - November 2025 imagine-ai.eu PROJECT OVERVIEW Interview With Gergely Sipos EGI Foundation, iMagine Project Coordinator imagine-ai.eu Watch the interview here youtube.com/watch?v=JVZb4eQk7OU iMagine Impact Report - November 2025 imagine-ai.eu 3 Introduction The iMagine project was launched under the Horizon Europe programme with the mission to revolutionise how researchers monitor and understand aquatic ecosystems through advanced artificial intelligence (AI) and open science. Running from September 2022 to August 2025, iMagine has brought together a diverse consortium of partners spanning environmental sciences, data management, AI developments, e-infrastructures and Research Infrastructure. Its overarching objective: to bridge the gap between marine and freshwater research communities and cuttingedge digital technologies, creating a robust, open-access platform for AI-powered image analysis that serves both scientific and societal needs. Through the development of new AI models, user-driven services, and a powerful cloud-based platform for AI model development, the project has empowered researchers to automate complex environmental assessments, accelerating insights and enabling real-time monitoring across freshwater and marine domains. Relevant Links iMagine Website https://www.imagine-ai.eu/ iMagine AI Platform https://go.egi.eu/imagine-aiplatform iMagine Zenodo Community https://go.egi.eu/imaginezenodo-community iMagine YouTube Playlist https://go.egi.eu/imagineyoutube-playlist Check out all the partners and RIs that made iMagine possible at page 12. 24 Partners 12 Research Infrastructures Supported 4 E-infrastructure Providers 17 Use Cases Supported 8 Key Exploitable Results 4,5M Funding iMagine Impact Report - November 2025 imagine-ai.eu 4 iMagine Internal Use Cases Marine Litter Assessment This use case utilises AI-driven image analysis methods to automatically detect and quantify floating marine litter from drone imagery, aiming to enhance monitoring efforts for reducing plastic pollution in aquatic environments. Oil Spill Detection This use case applied AI models to satellite imagery to detect and predict the spread of oil spills at sea, supporting rapid response efforts and minimising environmental damage from marine pollution incidents. Zooscan The Zooscan use case focused on applying AI techniques to classify and analyse zooplankton images captured by specialised imaging devices, to accelerate biodiversity assessments and ecosystem health monitoring. FlowCam Phytoplankton Identification The FlowCam use case developed AI algorithms to automatically identify and classify phytoplankton species from highresolution microscope images, aiming to streamline water quality assessments and ecological studies. EMSO Azores This use case leveraged AI-based video and image analysis to monitor deepsea hydrothermal vent ecosystems at the EMSO Azores observatory, aiming to automate the detection of biological and environmental changes over time. Underwater Noise This use case introduced AI methods to analyse hydrophone acoustic data for detecting and characterising underwater noise sources, aiming to support studies on the impact of human activities on marine ecosystems. EMSO OBSEA At the EMSO OBSEA coastal observatory, this use case developed AI models to analyse underwater imagery for tracking marine life activity, supporting longterm ecological monitoring and habitat assessment. Beach Monitoring The beach monitoring use case utilised AI models to automate beach seagrass wrack identification, shoreline extraction, and the detection of rip currents from beach imaging systems to improve coastal dynamics assessment and early warning systems development. Freshwater Diatom Identification This use case created AI models to classify freshwater diatom species from microscopic images, aiming to enhance the efficiency and accuracy of freshwater ecosystem biomonitoring programs. EMSO Smartbay Focusing on the EMSO Smartbay observatory, this use case developed AI models to process imagery and video from coastal waters and prawn burrow surveys. It aims to improve the automated detection of biological activity and develop models to assist in Prawn fisheries Surveys. ESMO Smart Bay also hopes to implement an existing Video Quality Assessment algorithm to review real-time and archived video data. iMagine Impact Report - November 2025 imagine-ai.eu 5 iMagine External Use Cases Sea Wave and Coastal Inundation Detection Methodology This use case introduced AI-based methods to detect sea wave patterns and coastal inundation events from sensor data and imagery, aiming to support early warning systems and coastal risk management strategies. Cold Water Coral Reefs Focusing on cold-water coral reef ecosystems, this use case used AI-driven image analysis to monitor reef health and detect changes over time, supporting biodiversity conservation efforts in deep-sea environments. DEAL The DEAL use case uses Swarm Learning to allow users to retain their image data and only share their models’ parameters, making it possible for users to efficiently collaborate and improve models with lower biases, while preserving data privacy. in iMagine, DEAL can compliment existing biodiversity capabilities while expanding the data available and allowing the team to verify and validate their cloud computing design requirements. AI-based Detection and Classification of Seafloor Litter This proof of concept evaluated the use of AI methods for the detection and classification of seafloor litter in deep-sea imagery retrieved from the Ifremer video archive and complemented with high-resolution mosaics produced during the 2023 and 2024 MOMARSAT cruises campaigns. Satellite Derived Bathymetry This use case applied AI techniques to satellite data to estimate underwater topography (bathymetry), aiming to provide faster and more cost-effective mapping of coastal and shallow marine areas. Fish Otoliths The fish otoliths use case developed AI tools to analyse the microscopic structures of fish ear bones (otoliths), aiming to automate species identification and support fisheries management and ecological research. EyeOnWater This use case enhanced the EyeOnWater citizen science platform by using AI to automatically validate smartphone images of water bodies, aiming to improve data quality and support large-scale environmental monitoring by the public. iMagine Impact Report - November 2025 imagine-ai.eu 6 Key Results Over the three years, the project has presented tangible instances and stories of success, bridged the gap between marine and freshwater research communities and cuttingedge digital technologies, and showcased the capabilities and potential of AI-powered image analysis to serve both scientific and societal needs. The project has contributed towards the Horizon Europe objectives with the following eight Key Results, KER8: Best Practices KER5: Flowcam Plankton Identification KER7: The iMagine AI Platform KER3: Marine Ecosystem Monitoring KER1: Marine Litter Assessment KER4: Oil Spill Detection KER6: Prototype Imaging Services KER2: Zooscan – Ecotaxa Pipeline iMagine Impact Report - November 2025 imagine-ai.eu 7 iMagine Impact New Products and Services iMagine AI Platform The project’s flagship product is the iMagine AI Platform, a cutting-edge computational platform designed to revolutionise image analysis in aquatic sciences. This platform was built as a generic, robust environment where users can develop, train, and share AI models for image processing at scale . It provides a one-stop solution covering the entire machine learning lifecycle – from data ingestion and annotation, through model training and validation, to deployment of models as ready-touse services. The iMagine Platform is implemented on a federation of cloud resources (leveraging four cloud providers from the EGI federation) with on-demand GPU and CPU computing and large-scale storage to meet the intensive requirements of AI processing . The platform is closely integrated with European-wide initiatives to maximise interoperability and user reach. As an open-access ecosystem, the iMagine platform will continue to enable researchers in aquatic science worldwide to develop, leverage and use advanced AI tools without needing to invest in their own compute infrastructure. This approach fosters collaboration and knowledge sharing across the community. The iMagine AI platform is expected to remain a lasting resource for the community, so that researchers across Europe can continue to access free at the point of use AI tools. AI Models and Services Building on this platform, iMagine developed a suite of AI services and models addressing specific scientific and environmental challenges. The project worked on over a dozen new image analysis services in collaboration with domain experts . These services cover a remarkably diverse range of aquatic science use cases, demonstrating the versatility of the platform. Together, this portfolio of AI services showcases how iMagine has operationalised AI for aquatic research. The project delivered several ready-to-use AI models which are usable and retrainable by others . In addition, the project created accessible AI services that allow end-users (such as marine researchers or environmental managers) to run complex image analyses on demand, without needing deep technical expertise . Other researchers and developers can build on these models and services to create domain-specific variants or adapt them for new environments. This facilitates the emergence of innovation cycles and derivative products. The project’s models and software code for services are openly licensed and can therefore act as a springboard for future innovation. SMEs, startups, and service providers in the blue economy, environmental consulting, and geospatial analytics can adapt iMagine’s services to deliver new commercial offerings without starting from scratch. 11,113,730 CPU-hours consumed 365,878 GPU-hours consumed 843 TByte Month storage delivered 18 AI Models Developed 7 Mature AI Services iMagine Impact Report - November 2025 imagine-ai.eu 8 Impact on Science Peer-reviewed Publications Peer-reviewed scientific publications are a fundamental output of high-quality research, representing the formal channel through which new knowledge is validated, disseminated, and embedded into the broader body of scientific understanding. They serve not only as a record of innovation but also as a foundation for cumulative discovery, often sparking new lines of inquiry, shaping methodologies, and influencing how future research is conducted. iMagine has generated a growing portfolio of peer-reviewed publications in high-impact journals and conference proceedings, spanning fields such as marine biology, remote sensing, computer vision, and environmental monitoring. These publications document the project’s methodological innovations, for example, the development of deep learning models for plankton and diatom classification, AI-based tools for marine litter detection, and automated monitoring frameworks for acoustic noise and underwater ecosystems. In many cases, the publications are accompanied by open datasets and code repositories, enabling other researchers to reproduce, validate, and extend the work. This integrated approach — sharing models, data, and the scientific rationale — enhances the credibility and utility of iMagine’s outputs. Publications also highlight collaborative, interdisciplinary approaches involving environmental scientists, data engineers, and AI researchers, showcasing how cross-sector collaboration can deliver real-world innovation. By presenting novel AI methodologies and applications for environmental research, iMagine publications will inform and inspire future studies. Researchers working in related domains can reference these works to build on validated approaches or propose follow-up investigations. As these publications are cited in new articles, conference papers, and theses, iMagine’s contributions will become embedded in the scientific canon, contributing to Europe’s standing in AI and environmental science research. Provision of Specifically Curated/ edited Data Provisioning of research data enables the research community, public and private entities to exploit these (digital) resources for their R&D or other purposes. These specifically curated datasets become a valuable resource to further develop products, innovations, studies, policies, etc. iMagine has prioritised open data from the outset, producing and publishing multiple curated and annotated imaging datasets tailored to aquatic science. These datasets are hosted in trusted open-access repositories such as Zenodo and are designed to be reused by both domain scientists and AI practitioners. These datasets have been formatted to meet FAIR standards, with comprehensive documentation, licenses, and persistent identifiers to support reuse and citation. Access to such clean, labelled data will lower the barrier to entry for new research. By openly sharing such large, domain-specific datasets, iMagine ensures that its scientific outputs can be diffused and reused by other researchers. These open datasets will also be valuable teaching tools in environmental science and AI curricula. Students and early-career researchers can use them in coursework or hackathons to gain practical experience with real-world data. Changing Fundamentals of Research Practice iMagine is a transformative research project that has not only created new knowledge but has also fundamentally changed the way research is conducted. The project has helped transform research practices in aquatic science. iMagine has contributed to a fundamental evolution in aquatic and environmental research practices by promoting the shift from manual and fragmented workflows towards AI-enabled, automated, and standardised approaches. Historically, tasks such as taxonomic identification, organism counting, and image segmentation, among others, relied heavily on time-consuming manual observation (e.g., delineation tasks) and the need for highly specialised personnel (e.g., ichthyologists). Data processing was often ad-hoc, non-standardised, and siloed within specific projects or institutions. iMagine has demonstrated that artificial intelligence, supported by open infrastructures, can automate and standardise these tasks at scale, drastically improving the efficiency, reproducibility, and objectivity of environmental monitoring. The project not only developed AI models and services but also systematically captured best practices, created training guidelines through its dedicated Competence Centre. This approach ensures that the knowledge is not locked within a single project participant but is accessible, reusable, and adaptable by researchers across Europe and beyond. By connecting imaging data processing directly with open science practices, iMagine has redefined how imaging data in aquatic sciences can and should be handled in the future. In doing so, it exemplified responsible, interoperable, and sustainable digital research practices, paving the way for broader adoption across disciplines. 16 Peer-Reviewed Publication 47 Conference Presentations 23 Datasets 3 Best Practice Documents 2.3M+ Labelled Images 1020 Views 990 Downloads iMagine Impact Report - November 2025 imagine-ai.eu 9 Creating and Shaping Scientific Networks iMagine has made significant contributions to building new research networks and reinforcing existing collaborations. From the outset, iMagine was conceived as a highly collaborative initiative that linked multiple research infrastructures and initiatives. It brought together aquatic biologists, data providers, developers and AI specialists across more than 20 partner institutes, forging a network that spans disciplines and countries. To further its impact, iMagine also ran an open call to attract new use cases from the marine domain. The project’s open call resulted in 6 new use cases, expanding not only its geographical but also its scientific reach. Moreover, iMagine connected with external projects and international initiatives to exchange knowledge and avoid siloing of results. Notably, it integrated efforts with European Open Science Cloud (EOSC) and the AI4EU (AI-onDemand) platform to ensure its solutions align with broader e-infrastructure efforts . iMagine has also acted as a customer validation channel and source of user requirements for the AI4EOSC platform. The project also partnered or shared outcomes with marine science networks like Blue-Cloud, data repositories like Zenodo and sister projects like ANERIS and AI4Life, multiplying the reach of its scientific innovations while providing an additional channel for exploitation of results coming out of these initiatives. These links mean that iMagine’s scientific advancements (data, tools, standards) are being considered and potentially adopted by other major initiatives. These linkages ensured that the project was not working in isolation but contributing to a pan-European innovation ecosystem. iMagine did more than deliver technologies. It has built bridges between people, institutions, and initiatives. These bridges will continue to serve as durable platforms for sustained scientific cooperation, enabling future co-creation, innovation, and policy influence. This enduring collaborative infrastructure ensures that iMagine’s impact will continue to grow, not only through its direct outputs but also through the vibrant and dynamic networks it has helped shape. iMagine has been collaborating with Zenodo under the HorizonZEN project, which supports EC programme beneficiaries to comply with the FAIR and open science requirements. Thanks to this collaboration, Zenodo has added extra elements to its metadata template to fit the aquatic science datasets. These changes will persist beyond the project and will make Zenodo even more attractive for publishing marine datasets in future.