CHOOSE DATA FOR THE TASK

Real places. Multiple views. Reconstructed geometry.

Explore captures as environments, not isolated 3D objects. Inspect the images alongside the spatial products retained from their reconstruction.

Browse images with reconstructed geometry →337 scenes match this file-presence filter.

A scene is more than a model file

A standalone mesh describes one representation of an environment. A scene dataset can also preserve the original photographs, camera-related files, orthomosaic and elevation products from the same capture. Those additional observations help you investigate how the model was reconstructed and whether it is useful for your own workflow.

The current collection contains aerial captures. Outdoor buildings, terrain and other environments must be assessed from the views actually present. A drone reconstruction should not be assumed to include building interiors, hidden surfaces, complete street-level detail or a watertight volume.

Choose by the reconstruction you need

The images-plus-geometry filter requires current-version source photographs and at least one retained point-cloud or mesh asset. A scene with only an orthomosaic does not satisfy that filter. File formats, counts and sizes remain visible on the scene page, so you can distinguish the available products before requesting access.

For novel-view or pose-conditioned work, add camera-data files to your search and inspect their contents. Camera file presence alone does not prove that every image has usable extrinsics and intrinsics. The reconstruction workflow, coordinate conventions and image associations need to be checked together.

Plan your own preprocessing

Inspect the imagery for overlap, exposure changes, motion blur and repeated textures. Inspect reconstructed surfaces for gaps, vegetation artifacts and missing viewpoints. Where a mesh includes textures, verify material references and the files needed to resolve them. Review the scene-specific quality notes instead of treating an export format as a quality grade.

This catalog does not promise prebuilt Gaussian splats, NeRF checkpoints, collision geometry or simulator imports. Source imagery and retained reconstruction files may be useful inputs to those workflows, but conversion, training and validation are separate work under the appropriate license.

Evaluate without confusing input and reference

A point cloud and mesh generated from the same photographs are related derived products. They can support consistency checks and workflow comparison, but they should not be presented as independent survey ground truth. Report that dependency when designing a benchmark.

Request the exact versions, formats and rights needed for your project. Evaluation samples are intended for internal inspection under the sample terms; a separate commercial agreement defines training, production and redistribution permissions.

FROM RESEARCH TO A SHORTLIST

Collections for this workflow

Browse scene-level selections with explicit requirements and live inventory counts.

All collections ↗
FROM THE CURRENT CATALOG

Inspect matching scene inventories.

These records match the required file types. Their suitability still needs evaluation.

All matching scenes ↗
United States / Arkansas

Wilmar drone capture · AD-29CEB70F

Retained together in this scene: 134 source image files, camera-data files, point clouds, orthomosaics.

Images: listed Camera data: listed 3D geometry: listed
6 retained data types1.3 GB Sample available
Israel / Center District

Even Yehuda drone capture · AD-D5149B24

Retained together in this scene: 100 source image files, camera-data files, point clouds, meshes, orthomosaics.

Images: listed Camera data: listed 3D geometry: listed
7 retained data types825.7 MB Sample available
United States / North Carolina

Asheville drone capture · AD-3D35708C

Retained together in this scene: 100 source image files, camera-data files, point clouds, meshes, orthomosaics.

Images: listed Camera data: listed 3D geometry: listed
7 retained data types1.6 GB Sample available