Three different kinds of camera information
Capture observations can include GPS positions, timestamps or gimbal orientation. Calibration describes the imaging model, such as focal parameters and distortion. Reconstructed camera poses describe estimated camera positions and orientations within a reconstruction. These are related concepts, but one does not substitute for another.
The catalog uses a camera-data asset category. It does not currently expose a per-scene certification that full intrinsics and extrinsics have been validated for every image. Results on this page are candidates for inspection, not a guaranteed pose benchmark or a promise of a particular calibration format.
Check the image-to-camera relationship
Open the camera files and locate the image identifiers. Confirm that those identifiers map to the source images you will actually use, including filename extensions, any renamed images and any excluded views. Check whether image resizing or cropping requires changes to the supplied intrinsic parameters.
Inspect the coordinate convention as well as the numbers. Determine whether transforms are camera-to-world or world-to-camera, which axes define the camera frame, how rotations are encoded, and whether translations are in a known unit. Geographic observations and reconstruction coordinates may use different reference frames.
Test projection consistency
For scenes with both camera data and reconstructed geometry, a small projection test can help reveal mismatched units, transform direction or image associations. Choose visible landmarks and inspect the result across several source views before scaling the ingestion workflow.
Such a check establishes internal consistency for your pipeline; it does not independently establish geodetic accuracy. The reconstructed poses and geometry may have been estimated together from the same image set. Read the listed alignment and accuracy notes before making external accuracy claims.
Request the fields, not just the category
In your inquiry, specify required intrinsic parameters, distortion model, pose convention, image identifiers, scene coordinate system and any minimum registered-image count. Include whether you need source observations, reconstruction outputs or both. This makes it possible to assess the retained files rather than assume a universal camera schema.
Evaluation samples vary by scene and may not include every camera-related file. Ask for clarification when the available sample is insufficient. Internal sample inspection is separate from licensed training, fine-tuning or production use.
