This paper presents initial results from the Artemis project, which develops digital infrastructure for the analysis of historical maps of the Scheldt River Valley. We focus on two computer vision tasks: parcel boundary segmentation and text recognition on cadastral maps. Our experiments show that fine-tuning is essential for historical materials, and that even imperfect segmentation results can support downstream tasks such as polygon extraction and spatial linking. The work highlights both the potential and challenges of applying computer vision methods in digital humanities contexts.