Abstract
The ability to quantify species distribution in shelly sediment is important to the geotechnical engineering community as it can be used to predict sediment strength under load. These sediments are comprised of many species of foraminifera, and each has unique structural properties that affect the sediment strength, making the classification of each individual particle paramount. This thesis proposes a method of classification using deep learning with convolutional neural networks. The methodology is built upon previous work and delves deeper into the taxonomy of foraminifera and bioclasts. Unlike previous work, the proposed approach is focusing on classifying individual species within higher order classes. Accuracies in the range of 86% are recorded with a custom-designed convolutional neural network. The performance of this architecture and its predecessors is tested against a dataset of three-dimensional X-Ray microtomography scans of offshore calcareous sand which contains the species studied in this paper.