I made this Python package for data scientists and researchers who are interested in applications of Graph Neural Networks (GNNs) to geospatial domains.
The point is, there are loads of different data sources that could be graphs (networks), such as streets, transit systems, mobility flows, and object proximities.
City2Graph can handle those different data sources in a standardised and integrated manner, converting into PyTorch Geometric tensors.
It also now supports rustworkx, which makes the feature engineering for network properties much faster by rust.
I’d love to hear any feedback or ideas of further improvements!
yutasato•2mo ago
The point is, there are loads of different data sources that could be graphs (networks), such as streets, transit systems, mobility flows, and object proximities.
City2Graph can handle those different data sources in a standardised and integrated manner, converting into PyTorch Geometric tensors.
It also now supports rustworkx, which makes the feature engineering for network properties much faster by rust.
I’d love to hear any feedback or ideas of further improvements!