Generative AI improves urban land-use planning with satellite images
Generative AI for urban planning: synthesizing land-use maps with satellite imagery via DualControlNet
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Traditional land-use models predict changes based on numbers but cannot create new land-use maps from simple descriptions. The authors developed DualControlNet, an AI model that generates detailed land-use maps and matched satellite images together, allowing better control and evaluation of outputs. They trained it using data from two cities and showed it produces more consistent and visually accurate results than earlier methods. This system can model different urban development scenarios by adjusting factors like building density based on text prompts.
What this means in practice
- •For urban planners: Generate high-resolution land-use maps and satellite images together to explore different development scenarios guided by text-based controls.
- •For geospatial analysts: Improve spatial consistency assessment by verifying generated land-use maps with corresponding satellite imagery for better map accuracy.
Authors
Qingqi Song, Yuzhou Chen, Lingqian Hu, Shenhao Wang
Abstract
Various land-use models characterize land-use patterns and facilitate scenario planning. However, existing models are largely predictive, forecasting transitions based on numeric data without synthesizing new land-use maps from simple text prompts. Meanwhile, generative AI in urban systems remains unimodal, weakly controllable, and evaluated mostly by generic image quality rather than domain-specific metrics. To address these gaps, we introduce DualControlNet, a generative land-use model that jointly synthesizes high-resolution land-use maps and satellite imagery, addressing multimodality, controllability, and verification challenges. Trained on over 40,000 paired tiles from Orlando and Chicago, DualControlNet is evaluated using land-use, spatial consistency, and image quality metrics. Relative to unimodal baselines, DualControlNet improves land-use diversity consistency by 62.8% and building coverage ratio consistency by 16.3%, more than doubles cross-modal spatial consistency, and triples visual quality. Scenario-based experiments demonstrate that DualControlNet supports sustainable development through multidimensional controls, increasing density and diversity, and responding to text prompts like "high-density development." Overall, this work advances land-use modeling via a multimodal, controllable, and verifiable generative framework, lowering technical barriers and enabling intuitive exploration of urban futures.