Machine learning applications to archaeological site detection: current approaches and future perspectives

Machine learning applications to archaeological site detection: current approaches and future perspectives

The last few years have seen an important increase of new AI-based methods for the detection of archaeological sites and features. Most case studies have focused on the application of CNN-based deep learning methods using lidar datasets. Although relatively efficient for the particular cases in which they have been applied, these perform object detection of specific spatial pixel patterns. However, most archaeological sites do not display any particular spatial pixel correlation. On the contrary, archaeological sites often come in a high variety of shapes and flavours. Some archaeological sites are, indeed, virtually invisible to current remote sensing technologies and the only way available to us to record them is traditional pedestrian survey. This talk will present current efforts by the Landscape Archaeology Research Group at the Catalan Institute of Classical Archaeology to develop semi-automated workflows for the large-scale detection of archaeological sites within a range of environments and visibility conditions using multiple sources.