Automating taxonomic and systematic search of benthic foraminifera in an online database

Abstract

Recent advances in the applications of deep neural networks in computer vision tasks such as image classification has seen a tremendous surge in interest. Several image classification algorithms can now be leveraged in automating some tedious tasks associated with benthic foraminifera research especially in sample picking, taxonomy and systematics. In this study, a small image identification model was built with 414 SEM micrographs representing twenty-one species of benthic foraminifera, using a convolutional neural network which achieved 84% model accuracy and 75% validation accuracy on previously unseen images. The model was also deployed through a web application to demonstrate how it may be useful in augmenting online databases such as the Ellis Messina catalogue and the World Register of Marine Species. These services although very valuable, can be modernized with image search functionalities to enhance their perpetual usefulness and continuity.