Transform geospatial relations into graph representations designed for spatial analysis and Graph Neural Networks (GNNs).
copied from cf-post-staging / city2graphcity2graph is a Python library for converting datasets of geospatial relations into graphs. It is designed to facilitate graph data analytics in particular for urban studies and spatial analysis. It supports PyTorch Geometric to enable graph representation learning, such as Graph Neural Networks (GNNs).
Key features include: - Construct graphs from morphological datasets (buildings, streets, land use) - Construct graphs from transportation datasets (public transport networks) - Construct graphs from contiguity datasets (land use, land cover, administrative boundaries) - Construct graphs from mobility datasets (bike-sharing, migration, pedestrian flows) - Convert geospatial data into tensors for graph representation learning - Support for various proximity methods (KNN, Waxman, fixed radius) - Integration with PyTorch Geometric's Data and HeteroData formats
The library enables urban researchers and data scientists to transform complex geospatial relationships into graph structures suitable for network analysis and machine learning applications.