Cartographic generalization is a fundamental task in cartography that involves producing smaller scale yet meaningful maps from large-scale spatial data. One of the most common techniques used to achieve this is simplification, which poses interesting challenges when it comes to automation. In recent years, deep learning models have been increasingly utilized in cartographic generalization re-search. Unlike traditional methods that rely on complex rules and thresholds, deep learning approaches are more flexible and can reduce the need for manual intervention, leveraging the vast amounts of training data from existing map series. This thesis aims to develop a deep learning model, using stable diffusion models, to generalize building and topographic maps.
Figure 1: Example of a prompt-based Map Generalization
Literature/references:
Feng, Y. (2023). Prompt-aided Map Generalization with Diffusion Models. In GIScience 2023 Workshop on CartoAI: AI for cartography.
Rombach, R., Blattmann, A., Lorenz, D., Esser, P., & Ommer, B. (2022). High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (pp. 10684-10695).
