Scale AV Perception Across Vehicle Platforms with NVIDIA Omniverse NuRec

Moving driving software to a vehicle with differently positioned cameras changes what its perception model sees. NVIDIA's NuRec workflow reconstructs recorded driving scenes in 3D and renders them from a target vehicle's camera positions and lens settings. Developers can use those synthetic views to prepare training data before a complete dataset from the new vehicle exists.

The reconstruction retains information such as object tracks and camera calibration, which can help adapt labels to the new views. A post-processing model called Harmonizer reduces rendering artifacts, but NVIDIA says it does not repair incorrect calibration, recover scene coverage that was never reconstructed, or replace validation of the target-camera views. Real driving data is still needed to ground and validate performance, and the post does not give measured results in its text.