News

2022-09

Organised by Rubens Lopes (a WWW PIC participant from Brazil) and other colleagues, under the auspices of the I/ITAPINA network, an online course on plankton imaging span the summer of 2022. On September 13th, Jean-Olivier Irisson, WWW PIC lead from France, presented a class on the use of machine learning for plankton and particle images. As part of the class, the tools developed in WWW PIC were demoed in front of the audience and, although live demos usually fail, everything worked perfectly. This class was a great opportunity to showcase our work and we hope many students will benefit from it. All classes are described and available online and the machine learning class is streamable too; you can skip to the demo of EcoTaxa and the demo of an unsupervised image plankton exploration and classification tool.

2022-06

Led by Thelma Panaïotis, PhD student in France, the WWW PIC team compared segmentation approaches for plankton images taken by shadowgraph instruments, in particular the In Situ Ichtyoplankton Imaging System. The comparison was published in Frontiers in Marine Science. We explored basic segmentation, Maximally Stable Extremal Regions, and Convolutional Neural Networks and discuss the merits and drawbacks of each. All approaches are implemented in open source software, made available in the infrastructure.

2022-01

A joint team of French and Japanese partners of WWW PIC published a comprehensive review of the applications of Machine Learning for the study of images of plankton or the detrital particles named "marine snow". The review was pushied in the 2022 edition of Annuel Reviews in Marine Sciences. It starts by exploring the history of the methods, then showcases how imaging combined with machine learning allowed to shed light on plankton ecoloy, and it ends with recommendations for the creation of a global network of experts and tools to process the flow of images prorduced by the current batch of instruments. The history described in the fist part is what led to the current design of the machine learning approaches in EcoTaxa. The final section reflects our experience building the WWW PIC infrastructure.