The third year of iMagine started with finishing the AI service development phase, then continuing with the delivery phase that enabled access to thematic AI image processing services across 5 thematic domains within aquatic sciences.
Objective 1: Deliver a scalable, shared IT platform for image analysis in marine and freshwater research.
The iMagine AI Platform, operational since December 2022 built on the AI4OS technology, successfully supported the 8 project use cases, and 7 external use cases. The use cases developed and shared 18 domain-specific AI modules via the platform.
Underpinned by four cloud infrastructure providers, the platform dramatically surpassed capacity delivery plans, providing 11.1 million CPU-hours and 365,878 GPU-hours for the use cases and their end users during the 3 years. All providers of the generic AI infrastructure setup are committed to continue resource provision after the project ends.
Objective 2: Advance existing image analytical services to increase research performance in aquatic sciences.
The 5 mature use cases of the project reached production stage in late 2024, and started the cross-national delivery of seven image analytical services with Virtual Access. These services were deployed on the iMagine platform (Objective 1), and was complemented with support for user uptake in the form of conference presentations, scientific papers, information sharing webinars and technical user support.
Users gained access to the services via an online form, with access options ranging from ‘Quick try out', full-scale 'Model Inference/Analyse runs’, 'Model re-Training' and ‘custom local deployment' options. In the final project year the thematic services served 238 users. Besides sharing the services, all related software code and the labelled training datasets have been also made available as open-source resources.
Objective 3: Develop prototype new image analytical services and datasets that can accelerate progress towards healthy oceans, seas, coastal and inland waters
The three project prototype use cases completed model training and published their services on the iMagine platform as Minimal Viable Products (MVPs) for validation. The 8 iMagine use cases made significant contribution to AI image publishing: 23 datasets have been published via Zenodo and other domain-specific repositories, including over 3 million open access images, with 2.3 million of them annotated for supervised AI model training.
Objective 4: Capture and disseminate development and operational best practices to imaging data and image analysis service providers
The project successfully captured and disseminated best practices along the entire AI model development-training-validation-delivery pipeline, establishing key collaborations for long-term impact. Three best practices documents were published in Zenodo, and have been downloaded over 900 times in total.
Key collaboration with the Zenodo repository, via the Horizon-ZEN project, resulted in a new metadata template that better supports the description and discovery of aquatic science datasets, ensuring legacy well beyond the project’s duration.
Objective 5: Deliver a portfolio of scientific image and image analytics services targeting researchers in marine and aquatic sciences
A robust service portfolio was delivered and supported by continuous expert collaboration. The portfolio now includes a generic AI platform, seven services from the five mature use cases and three MVPs from the prototype use cases. All related models, code, and training data are available as open-source components.
