Spatial dataset collections.
Start with the representations your workflow needs, or the environments it needs to see. Every collection draws from the same reviewed scene catalog.
Choose a representation set.
Multimodal spatial scene datasets
Browse real-world scenes with source images, camera-data files and reconstructed geometry. Compare current inventories for spatial AI research and licensing.
Multi-view image datasets with camera-data files
Find scenes retaining multiple source images and camera-data files. Inspect image associations, intrinsic parameters and pose conventions for view-based research.
Image and point-cloud scene datasets
Browse scenes containing source images and current photogrammetric point-cloud files. Compare reconstruction context, retained attributes and licensing options.
Source-image and 3D mesh scene datasets
Explore real-world scenes with source photographs and retained 3D mesh files. Assess surface coverage, textures and reconstruction limitations before licensing.
Choose an environment.
Built-environment 3D scene datasets
Browse reviewed built-environment scenes with source imagery and reconstructed geometry. Compare buildings, outdoor structure and aerial-view coverage for 3D research.
Construction and infrastructure spatial datasets
Explore reviewed construction, infrastructure and industrial scenes with images and geometry. Inspect capture dates, retained products and site-specific limitations.
Terrain and vegetation 3D scene datasets
Browse reviewed terrain, agriculture and vegetation scenes with imagery and geometry. Assess surface coverage, reconstruction artifacts and environmental diversity.
Purposeful groups,
not duplicated datasets.
A scene can match more than one collection. Counts describe membership, not additional unique captures. Every result must pass the shared publication checks and the collection’s stated file and category requirements.
Representations are checked against current, non-sample assets. Camera-file presence does not certify poses; image-derived geometry is not independent survey ground truth. Each collection explains what still needs evaluation.
How to select spatial training data →Need a narrower collection?
Specify your required files, environments, scale and intended use.