Classifying Pottery Fragments by Decorative Technique Using a Convolutional Neural Network

This project presents and tests a method of classifying ancient Greek pottery fragments based on exterior decorative techniques using deep learning. Classification using a convolutional neural network can potentially help save hours of work for archaeologists who can use that time to do higher-level pottery and context analysis. Although this project focuses on two well-known decorative techniques called black-figure and red-figure, I believe this method can be applied to other decorative techniques as well.  I discuss the challenges I encountered during the process and explain how this method can be improved to generate greater success. Finally, I suggest what future work should be done to turn this method into a user-friendly, effective tool for archaeological research.