Applying automated object detection in archaeological practice: A case study from the southern Netherlands

Within archaeological prospection, Deep Learning algorithms are developed to detect objects within large remotely sensed datasets. These approaches generally tested in an (ideal) experimental setting but have not been applied different contexts or ‘in the wild’, that is, incorporated prospection. This research explores applicability, knowledge discovery—on both a quantitative and qualitative level—and efficiency gain resulting from employing automated detection tool called WODAN (Dutch) practice. has used barrows Celtic fields LiDAR data Dutch Midden-Limburg area, which differs archaeology, geo-(morpho)logy land-use Veluwe it was developed. The results show able potential fields, including previously unknown examples, provided information about structuring of landscape past. Based on results, combined human-computer strategies argued, complementary, rather than substitute role, manual analysis. can offset inherent biases analysis deal with problem current methods only similar pre-defined target class(es). incorporation into highlight areas interest enhance add detail existing predictive maps, seems logical feasible.