FRUC
Scene reconstruction · Bundled collaborative driving scenes
About FRUC
FRUC reconstructs dynamic driving scenes from views captured by collaborating agents. The demo uses bundled V2X-Real samples: an ego vehicle and a collaborator, each observed at two consecutive timestamps.
The model predicts cameras and a Gaussian scene representation from the images. You can inspect a new viewpoint, compare collaborative fusion with an ego-only reconstruction, or remove dynamic Gaussians. These controls make the Space useful for examining the contribution of extra viewpoints.
How to use FRUC
- Choose a bundled cooperative driving scene.
- Select the reconstruction and camera-movement settings, then run the demo.
- Compare fusion settings or dynamic-object removal in the rendered result.
What to know before you start
- The documented workflow uses bundled scenes rather than a general-purpose video upload.
- Removing dynamic Gaussians is a scene-editing demonstration, not a validated reconstruction of every hidden surface.
The provider controls availability, sign-in, processing, and downloads. Save any result before reloading or closing the demo.
Do I need camera calibration files to try this demo?
No. The documented workflow estimates camera parameters from its input RGB views and provides the example scenes for you.
