City2Graph · Export for GeoAI

Export a Burjassot graph for graph neural networks: JSON or GraphML, homogeneous or heterogeneous, plus a standalone PyTorch Geometric loader script.

This walkthrough mirrors the Export panel.

Before you begin

Have a graph object active — a street network, a proximity graph, a morphological graph, or a metapath result.

1. Choose the format and graph type

In Export, set:

  • Format — JSON or GraphML.
  • Graph Type — Homogeneous (single node/edge type) or Heterogeneous (multiple layers, e.g. a morphological or metapath graph).

Selecting the export format and graph type.

2. Export the graph

Set the Export Path and press Export Graph. The file is written in a layout city2graph.gdf_to_pyg can load into a PyTorch Geometric Data or HeteroData object.

Exporting the graph to disk.

3. Save the loader script

Press Save Loader Script to write a self-contained Python file that reloads the exported graph into PyTorch Geometric without Blender — ready to drop into a training pipeline.

Saving the standalone PyG loader script.

Next steps

This completes the panel tutorials. For full end-to-end scenarios, see the Examples.

Back to top