A Radiolarian Classifier using Convolutional Neural Networks

Radiolarians are zooplankton that drift around the oceans and settle to the bottom of the ocean after their death. Because of their diversity, abundance, and their existence since the Precambrian period, they became an important fossil for agedating, correlation, and studying deep-sea sediments. Existing traditional methods of identifying radiolarians are based largely on the study of skeletons and cross-referencing with known species, and were proven very difficult. In this study, we propose a method on classifying radiolarians skeletons with the use of Convolutional Neural Networks. We have covered both cases of radiolarians from different genera classes and from different species within a genus class, to cover both the breadth and depth of this organism's diversity.