Generative AI for Urban Planning: Synthesizing Satellite Imagery via Diffusion Models

TL;DR

This work introduces a diffusion-model-based framework, augmented with ControlNet, to synthesize high-fidelity satellite imagery conditioned on land-use descriptions and OpenStreetMap-derived constraints for urban planning. Using Mapbox satellite tiles and OSM layers across Chicago, Dallas, and Los Angeles, the authors build a data pipeline, fine-tune stable diffusion, and evaluate both quantitatively (FID/KID) and through large-scale user studies. They demonstrate cross-city texture and layout transfer, adherence to constraints, and diversity of designs under identical inputs, enabling rapid visualization and exploration in planning workflows. The results indicate strong fidelity and broad acceptance among non-expert viewers, highlighting the potential of GenAI to augment planning communication, public engagement, and design ideation while outlining avenues for richer inputs, multi-scale coherence, and equity-focused future work.

Abstract

Generative AI offers new opportunities for automating urban planning by creating site-specific urban layouts and enabling flexible design exploration. However, existing approaches often struggle to produce realistic and practical designs at scale. Therefore, we adapt a state-of-the-art Stable Diffusion model, extended with ControlNet, to generate high-fidelity satellite imagery conditioned on land use descriptions, infrastructure, and natural environments. To overcome data availability limitations, we spatially link satellite imagery with structured land use and constraint information from OpenStreetMap. Using data from three major U.S. cities, we demonstrate that the proposed diffusion model generates realistic and diverse urban landscapes by varying land-use configurations, road networks, and water bodies, facilitating cross-city learning and design diversity. We also systematically evaluate the impacts of varying language prompts and control imagery on the quality of satellite imagery generation. Our model achieves high FID and KID scores and demonstrates robustness across diverse urban contexts. Qualitative assessments from urban planners and the general public show that generated images align closely with design descriptions and constraints, and are often preferred over real images. This work establishes a benchmark for controlled urban imagery generation and highlights the potential of generative AI as a tool for enhancing planning workflows and public engagement.

Figures (12)

• Figure 1: The proposed generative urban design workflow
• Figure 2: Land use, transportation, and waterway layers from OpenStreetMap
• Figure 3: The selected study areas around Los Angeles, Dallas, and Chicago
• Figure 4: Illustration of a training pair
• Figure 5: ControlNet Architecture