Dataset Viewer
The dataset could not be loaded because the splits use different data file formats, which is not supported. Read more about the splits configuration. Click for more details.
Couldn't infer the same data file format for all splits. Got {NamedSplit('train'): ('imagefolder', {}), NamedSplit('validation'): ('json', {})}
Error code:   FileFormatMismatchBetweenSplitsError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Spatial-World showcase assets

Visualization tiles for Spatial-World, covering perception, dense reconstruction, spatial reasoning, and world simulation. Scenes were picked for color and for indoor, architecture, nature, street, night, and stylized settings.

Layout

Path Count What is inside
01_dense_geometry/tiles 60 scenes RGB, diagonal / latitude / up-vector fields, raw depth, glow depth, normals, point cloud
02_world_trajectory/tiles 50 clips Anchor frame, predicted and reference frames, trajectory box
03_camera_understanding/tiles 60 scenes Direct camera fields, tilted input, leveled closed-loop view, height class in record.json
04_spatial_vqa/tiles 75 questions Image, question card, prediction record
contact_sheets 6 sheets Grids for browsing each task

raw_model/ holds the model outputs used to render the tiles. dense_5x.json, world_5x.json, and vqa_5x.json are the scene lists.

Source photographs come from the project examples and from NYUv2, WorldScore, RealEstate10K, Puffin, WorldModelBench, and BLINK. Those images keep their original dataset licenses. The overlays, point clouds, and trajectory renders are model outputs for this paper.

Camera height labels in this batch: eye-level 29, low-position 16, high-position 7, aerial 6, underwater 2. Spatial VQA on this batch: 54/75 correct.

Downloads last month
143