SnowVision, named after Swift Creek researcher Frankie Snow and the Computer Vision field, features unique, state of the art image matching software to study curve patterns from existing pottery fragments. With start-up funding from the National Center for Preservation Technology and Training (NCPTT), the University of South Carolina’s (USC) project team developed a prototype of SnowVision. With NCPTT funding, Dr. Wang, who leads the Computer Vision part of the USC team, and his graduate students tested a series of curve matching metrics and then developed and refined a pattern decomposition algorithm, choosing the metrics and algorithm that performed best on our subject matter. Project code generated during the above grant period has been published on GitHub (https://github.com/SnowVision/SnowVision1.0) and documented in a peer-reviewed publication in the Journal of Electronic Imaging (Zhou et al. 2017; see also Yuhang et al. 2018). The success of this prototype greatly aided the USC team in securing additional funding from NSF to further refine the software.
Today, dozens of archaeological collections hold tens of thousands of Swift Creek sherds from sites in Georgia, South Carolina, Alabama, and Florida. Many of these collections have never been studied for Swift Creek designs and design matches. Digitizing the sherds using a 3D scanner is an essential part of the project, providing primary data for software development. To date, the team has completed the digitization (3D scanning) of 3,000 pottery sherds curated at Georgia Southern University’s R. M. Bogan Archaeological Repository, a major repository of Swift Creek collections. These data have proven to be a critical component of the development of an automatic method to extract curves on sherds and to identify underlying designs on sherds with single and composite or “overstamped” patterns.
Current project goals include integrating the SnowVision software with a project website accessible to researchers and the public.