Datasets:
Tasks:
Image-to-3D
Modalities:
Geospatial
Languages:
English
Size:
100K<n<1M
Tags:
3d-point-cloud
point-cloud-generation
city-scale
remote-sensing
satellite-imagery
digital-surface-model
License:
Pipeline scripts
These scripts reconstruct City3D-MultiGen from the source data, organized by city. Only the canonical (non-duplicate) version of each stage is kept.
melbourne/
Processing for the City of Melbourne 3D Point Cloud.
| Script | Role |
|---|---|
export_las_blocks_noKML.py |
The full tiler: partitions the source LAS into 150 m tiles and produces the per-tile point cloud, DSM, and BEV render. |
grid_from_kml.py |
Builds the tile grid (metadata/*_grids.json) from the KML tile index. |
Obtain_corresponding_map_signed.py |
Fetches satellite + styled map from the Google Maps Static API and parses the per-class semantic masks. Reads GOOGLE_MAPS_API_KEY and GOOGLE_MAPS_URL_SIGNING_SECRET from environment variables. |
holicity/
Processing for HoliCity (London) FBX meshes.
| Script | Role |
|---|---|
export_las_blocks_noKML.py |
Samples point clouds from the meshes and tiles them (same full tiler as Melbourne). |
convert_coord.py |
Converts HoliCity local coordinates to geographic coordinates. |
add_coord_head.py |
Writes the geographic coordinate header onto each tile. |
check_coord.py |
Sanity-checks the coordinate alignment of generated tiles (optional utility). |
Obtain_corresponding_map_signed.py |
Fetches satellite + semantic maps (same Google API, your own key). |
build_dataset.py
Top-level orchestrator: tiling → map fetching → DSM/BEV, assembling the per-tile layout described in the top-level README.
⚠️ Before running
- Set your Google credentials as environment variables (see top-level README).
- API keys/secrets have been removed from these scripts and are read from the environment. Do not re-introduce hardcoded credentials — this repo is public.
- Edit the input/output paths at the top of each script to match your local layout.