Ceramics are used to address issues of chronology, craft production, and socio-political organization. However, it can take years to fully document and process large collections. Using a ceramic typology for the Early Bronze Age developed for the ARCANE Project in the Northern Levant as a case study, CRANE researchers are creating an automated ceramic shape recognition tool.
Currently, researchers identify ceramics by visually matching sherds against similar drawings. The shape recognition algorithm will expedite this process by matching 2D digitized images against the ceramics in the database. The future goal is to fully automate the process by scanning a sherd with a computerized scanner, creating an image that would then be put through the algorithm.
Lynn Welton is collaborating with Eugene Fiume’s computer science team at the University of Toronto on the project. The ultimate objective is to examine distribution patterns of third millennium ceramics throughout the Orontes Watershed in order to help validate or reject existing concepts of ceramic ‘provinces’ or ‘cultures’.
This research will then be expanded by incorporating the typologies for the Middle and Late Bronze Ages (based on work by CRANE Collaborator Marco Iamoni) and the Iron Age (based on work by CRANE Collaborator Matt Whincop). It will be used to examine relations between sites within the CRANE research environment in ways that were only possible in the past through exhaustive and time consuming research.
In 2015, the Pottery Informatics Query Database (PIQD) project joined CRANE. PIQD was developed by a research team at the University of California San Diego led by Neil Smith, now based at King Abdullah University of Science and Technology. The Pottery Informatics Query Database, on a limited basis, allowed users to identify, compare and create typologies from pottery fragments based on their shape. It focused predominantly on the pottery data from Israel.
PIQD has been integrated into the ceramic Repository (know as CROW - Ceramic Repository of the Orontes Watershed), and the database is continually being added to as excavations continue at some sites, or new projects are added. Presently, the team is continuing to develop the shape recognition algorithm using different approaches than initially proposed by CRANE or PIQD. Andy Chow (U of T Computer Science) will be working on the project as part of his doctoral research. See Shape Matching Project.