The official code of Remote Sensing of Environment paper Learning to Reason over Multi-Granularity Knowledge Graph for Zero-shot Urban Land-Use Mapping.

Abstract: This paper introduces a multi-granularity knowledge graph reasoning (mKGR) framework. Only with indirect supervision from other tasks, mKGR can automatically integrate multimodal geospatial data as varying granularity entities and rich spatial-semantic interaction relationships. Subsequently, mKGR incorporates a novel fault-tolerant knowledge graph embedding method to establish relationships between geographic units and land-use categories, thereby reasoning land-use mapping outcomes. Extensive experiments demonstrate that mKGR not only outperforms existing zero-shot approaches but also exceeds those with direct supervision. Furthermore, this paper reveals the superiority of mKGR in large-scale holistic reasoning, an essential aspect of land-use mapping. Benefiting from mKGR's zero-shot classification and large-scale holistic reasoning capabilities, a comprehensive urban land-use map of China is generated with low-cost.

• Products: Publicly accessible on ArcGIS Online, download the products on Zenodo.
• Code: Publicly available in this repository.
• Dataset: Publicly available on Zenodo.
• Docs: Apologies, but due to other recent projects, the detailed documentation will be provided in October. If you encounter any issues during reproduction, please raise an issue or send an email directly.

Ubuntu 20.04 (or other Linux distribution), one GPU (video memory greater than 12GB and support cuda)

• python>=3.11.5
• numpy>=1.26.2
• pytorch>=2.2.1
• pandas>=2.2.2
• geopandas>=0.14.0

Option 1: Directly download the constructed graph.

Option 2：Construct the graph in the KG_construction folder.

Train the graph embedding in the KG_embedding folder and infer to obtain the land-use mapping result.

figure_script: The code for generating the figures in the paper.

landuse_app: The code for 15-minute city application of land-use mapping results.

We have published the land-use mapping and 15-minute walkability results of China on ArcGIS Online.

If you have any questions about it, please let me know. (Create an 🐛 issue or 📧 email: [email protected])