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:
Tile metadata
Per-tile geographic coordinates are regenerated by the pipeline rather than shipped as a static file. (An earlier exported grid file did not match the released tile set, so it was removed to avoid confusion.)
Files
| File | Description |
|---|---|
melbourne_tile_index.kml |
Tile index for Melbourne, viewable in any GIS tool; input to scripts/melbourne/grid_from_kml.py. |
splits/ |
Train / validation / test tile-ID lists. Generate with scripts/make_splits.py (see splits/README.md). |
Regenerating the tile grid
The tiling scripts derive each tile's geographic extent directly from the source LAS, so no pre-computed grid file is required. If you want an explicit grid as JSON, run:
python scripts/melbourne/grid_from_kml.py # writes output_grids.json from the tile index KML
Each entry provides the tile id, grid row/column, and geographic bounding box (UTM + WGS84).