Project Overview
RLEA is developing the application of machine learning (ML) algorithms to identify phytoliths. These robust microscopic silica ‘casts’ of plant-cells are a major component of many archaeological and paleoenvironmental deposits. Currently, studying phytoliths is very time-consuming, but the researchers at the RLEA are developing deep learning machine learning algorithms that can analyze large amounts of data quickly. This will allow them to study many more ancient sites and understand how people used plants and impacted their environments in the past.
Related publications
Berganzo-Besga, Iban, Hector A. Orengo, Felipe Lumbreras, Paloma Aliende, and Monica N. Ramsey. 2022. Automated Detection and Classification of Multi-Cell Phytoliths Using Deep Learning-Based Algorithms. Journal of Archaeological Science 148:105654. https://doi.org/10.1016/j.jas.2022.105654.
Berganzo-Besga, Iban, Hector A Orengo, Felipe Lumbreras, and Monica N Ramsey. 2025. Deep Learning Black Box and Pattern Recognition Analysis Using Guided Grad-CAM for Phytolith Identification. Annals of Botany:mcaf088. https://doi.org/10.1093/aob/mcaf088.
Ramsey, Monica N, Melanie Pugliese, Lachlan Kyle-Robinson, Iban Berganzo-Besga, Rebecca Roberts, Jennifer Bates, Francesca D’Agostini, et al. In review. Experts Against Automation? Comparing Artificial Intelligence and Human Identifications of Phytoliths.
Images
