Moving in Space and Time.

We know surprisingly little about the historical transport infrastructure in large parts of Austria. There is some knowledge of the basic Roman transport infrastructure, and at the local level people sometimes have an amazing understanding of the remains of pre-modern roads, which sometimes still play a role, especially in rural areas. However, we often lack a broader understanding of pre-modern transport infrastructure. Between 1817 and 1861, Habsburg Central Europe was surveyed in the Franziszeische Kataster, a high-resolution cadastral map containing all the road infrastructure in Styria in the 1820s. We plan to develop a tool that extracts this information on a large scale - automatically, using appropriate deep learning tools that have only recently become available. Manually extracting this data from an estimated 10,000 map sheets (each covering 2 km) would take a highly skilled human team decades to complete. The team we are assembling for this task consists of historians, archaeologists, computer and data scientists from the University of Graz and the Graz University of Technology. This group has already successfully completed a similar task and will now attempt to take this endeavour to a new level. At the heart of this initiative is the development of two innovative artificial intelligence tools in two successive steps. In a first step, we develop PATHFINDER, which is designed to scan historical maps and high- resolution aerial photography to detect remnants of long-lost infrastructures. We use state-of-the-art deep learning technology to realize this approach. Pathfinder will make use of so-called CNNs (convolutional neural networks) and vision transformers, to scan the sources that contain the information we are looking for and to detect and extract the target data automatically upon its completion. It will then generate a GIS-layer, which can easily be integrated in research operations that require this data. In the second step, PATHMAKER will be tasked with reconstructing possible networks based on these findings. PATHMAKER will be able to look at incomplete networks of historical traffic infrastructure and to make suggestions for the reconstruction. It will learn and apply the logics of Roman and medieval roadmaking, for instance, and it will be able to decide when to apply the one and when to apply the other. The project sets a high bar for originality and innovation, bringing state-of-the-art deep learning models to bear on historical map analysis. The combination of AI with historical research promises new insights into the design logic of ancient road architects and offers a comprehensive understanding that could redefine the history of the region.