A multigenerational workflow: Applying Deep Learning tools on old maps to detect near-invisible historic sites.

Summary

This is an abstract from the "Practice, Theory, and Ethics of Machine Learning in Archaeology" session, at the 90th annual meeting of the Society for American Archaeology.

While archaeologists aim to use the latest technology to detect, classify, or analyze archaeological sites, they still face the classic problem that some sites are simply no longer visible due to soil deposition and erosion. While satellite imagery and aerial LiDAR data can sometimes help us see the outlines of certain buried archaeological features, sites covered by thick layers of sediment or that have been anthropogenically modified can still evade us. This is where going back to old data sources such as plat or general land office (GLO) survey maps can prove useful. In most cases where they are available, plat or GLO maps preserve information on land ownership, homesites, or other human-made features that were deemed important by original surveyors and can still be useful during modern archaeological surveys. Unfortunately, most of those maps are not georectified or vectorized, and thus cannot easily and quickly be used to pinpoint site locations on the landscape. Here, we present our workflow to train Machine Learning models on old georectified survey maps to extract the location of georectified homesites and legacy roads and thus create a GIS resource that can help cultural resource managers better document and protect their land.

This Resource is Part of the Following Collections

• Society for American Archaeology 90th Annual Meeting, Denver, CO (2025)   •
• Practice, Theory, and Ethics of Machine Learning in Archaeology

Keywords

Record Identifiers

Abstract Id(s): 51931