NEXT   CONTENTS   SUMMARY   ISSUE   HOME 

Classifying and Visualising Roman Pottery using Computer-scanned Typologies

Jacqueline Christmas1 and Martin Pitts2

1. Department of Computer Science, University of Exeter. Email: J.T.Christmas@exeter.ac.uk
2. Department of Classics and Ancient History, University of Exeter. Email: M.E.J.Pitts@exeter.ac.uk

Cite this as: Christmas, J. and Pitts, M.E.J. 2018 Classifying and Visualising Roman Pottery using Computer-scanned Typologies, Internet Archaeology 50. https://doi.org/10.11141/ia.50.14

Summary

A plot of outer circularity versus width/height ratio, where each point is represented by the cross-section outline of the pot so that we can see how the shape varies across the group

For many archaeological assemblages and type-series, accurate drawings of standardised pottery vessels have been recorded in consistent styles. This provides the opportunity to extract individual pot drawings and derive from them data that can be used for analysis and visualisation.

Starting from PDF scans of the original pages of pot drawings, we have automated much of the process for locating, defining the boundaries, extracting and orientating each individual pot drawing. From these processed images, basic features such as width and height, the volume of the interior, the edges, and the shape of the cross-section outline are extracted and are then used to construct more complex features such as a measure of a pot's 'circularity'. Capturing these traits opens up new possibilities for (a) classifying vessel form in a way that is sensitive to the physical characteristics of pots relative to other vessels in an assemblage, and (b) visualising the results of quantifying assemblages using standard typologies.

A frequently encountered problem when trying to compare pottery from different archaeological sites is that the pottery is classified into forms and labels using different standards. With a set of data from early Roman urban centres and related sites that has been labelled both with forms (e.g. 'platter' and 'bowl') and shape identifiers (based on the Camulodunum type-series), we use the extracted features from images to look both at how the pottery forms cluster for a given set of features, and at how the features may be used to compare finds from different sites.

Go to article Table of Contents.

AHRC logo