Freshwater diatoms identification

Setting up a development and training environment at the iMagine platform

Building a database for inserting and extracting of diatom images and gathering from various sources an extensive quality-controlled dataset of labelled (species name, traits estimation) light microscopy images for training and testing the CNNs

Analysis and training of AI models (CNN) fit for automatic classification of diatoms

Developing a GUI, using the iMagine platform, for user interaction

Documenting approach and experiences with resulting prototype

Contributing to dissemination and outreach, especially towards stakeholders (e.g., water agencies) and teachers (incl. for private consultancies)

Image database

Diatoms are unicellular microalgae present in all aquatic environments. They are routinely used as bioindicators for the ecological diagnosis of inland waters (rivers, lakes) as part of the implementation of the EU Water Framework Directive (WFD; Directive 2000/60/EC).

The development of high-throughput analyses will improve the diagnostic tools currently available, notably by improving the sampling effort, extending the current monitoring network to a larger number of stations and generating AI-based new metrics.

 These improvements will have great potential all over Europe and foster the development of new AI-based bioindicators targeting other ecosystems (diatoms are present in all aquatic environments) and key bioindicators (e.g., benthic invertebrates).