This project draws on technological advances in the field of computer vision to address logistical challenges and time constraints in the field of archaeology. At USC, SnowVision represents a multi-year collaboration between three units in two colleges. The software program will enable researchers to achieve in days and weeks what has literally taken individuals -- such as Frankie Snow, for whom the software is named -- years and decades to achieve. In archaeology, SnowVision will advance research topics related to geographical mobility, the materiality of pottery, and social learning in foraging societies of the southeastern Woodlands. In computer vision, SnowVision is advancing research in curve segmentation and 3D-image extraction.
The subject matter of SnowVision is a kind of decorated pottery made between A.D. 100 and 800 by Native Americans living in the lower Coastal Plain of the southeastern US. The pottery designs, comprised of often-combined elements including circles, teardrops, and figure 8s, were initially carved onto wooden paddles and subsequently transferred to pottery vessels through the manufacturing technique of ‘coiling with paddle-and-anvil’. The wooden paddles carry additional information that is also transferred to vessels. For example, although the designs themselves are largely unique, the paddles may also have carving or design flaws that mark them like twists or turns of the lines in a fingerprint. Unfortunately, the paddles have not survived, but the information on them is preserved on the surfaces of pottery sherds -- the most abundant artifact type for the time period.
Archaeologists who study this pottery recognized long ago that the fingerprint-like paddle designs could be tracked across multiple habitation sites, making it possible to address questions of human movement, for example, on a regional scale with almost-unprecedented granularity. SnowVision will dramatically increase the speed with which researchers are able to assess matches between sherd and design and will incentivize laboratories to digitize their collections of related subject matter, resulting in a boost of new data on paddle matches.
The SnowVision project is supported by grants from the National Park Service’s National Center for Preservation Training and Technology under Grant No P16AP00373 (2016-17) and the National Science Foundation under Grant No. 1658987 (2017-19)