Computer Vision Best Practices in Computational Archaeology

Summary

This is an abstract from the "Interdisciplinary Approaches to Landscape Archaeology - Part 2" session, at the 90th annual meeting of the Society for American Archaeology.

Landscape archaeology has progressed enormously in recent years thanks to the introduction of computer vision (CV) new technologies (Argyrou and Agapiou, 2022). Besides, Machine Learning (ML) has demonstrated its application to other archaeological fields beyond site detection (Berganzo-Besga et al., 2021), for example, the identification of recovered data, such as geochemical analysis (Oonk and Spijker, 2015) phytolith detection and classification (Berganzo-Besga et al., 2022), and the identification of ceramic fragments (Wright and Gattiglia, 2018) between others.

This lecture is intended to be a guide for the application of ML in archaeological research through a list of best practices. Both training and validation approaches will be presented. To do this, the most common problems encountered by archaeologists will be taken into account, such as the low-density of archaeological features to be detected and the small amount of training data available (Berganzo-Besga et al., 2023). Likewise, a series of methods will be shown to deal with the high number of false positives present in the algorithms.

The main objective of this talk is to present a comprehensive guide on the design, application and validation of CV methods, the most applied technology within the field of computational archaeology, for automated archaeological features identification.

This Resource is Part of the Following Collections

• Interdisciplinary Approaches to Landscape Archaeology - Part 2   •
• Society for American Archaeology 90th Annual Meeting, Denver, CO (2025)

Keywords

Record Identifiers

Abstract Id(s): 52647