Dataset Viewer
Auto-converted to Parquet Duplicate
preview_image
imagewidth (px)
800
1.92k
environment
stringlengths
12
136
scene_id
stringlengths
33
33
textured_usd
stringlengths
72
72
⌀
base_glb
stringlengths
55
72
source_code
stringlengths
110k
222k
isaac_sim_export
stringclasses
1 value
mujoco_export
stringclasses
1 value
gazebo_export
stringclasses
1 value
inputs
stringlengths
61
78
generation_model
stringclasses
1 value
part_count
int64
145
976
joint_count
int64
0
69
parts
stringlengths
60
77
articulation
stringlengths
67
84
measurements
stringlengths
67
84
physics
stringlengths
62
79
materials
stringlengths
64
81
mujoco_mjcf
stringlengths
67
84
gazebo_sdf
stringlengths
61
78
category
stringclasses
3 values
mesh_in_input
bool
2 classes
isaac_usd
stringlengths
68
85
School entrance and street
135yCY9mek1lgvxr6K-chsrWBQKLNlWbZ
scenes/135yCY9mek1lgvxr6K-chsrWBQKLNlWbZ/simulation/environment_pbr.usda
scenes/135yCY9mek1lgvxr6K-chsrWBQKLNlWbZ/base/model.glb
"""PS-IS 229 street frontage: an outdoor digital twin of a New York City residential block. BRIEF R1 Faithful, editable, simulation-ready twin of the photographed OUTDOOR place (8 photos + 456-frame 3 fps walkthrough, intrinsics f=400 px on 800 px => 90 deg hfov). R2 Reconstruct layout, ground, buildings, ...
✓
✓
✓
scenes/135yCY9mek1lgvxr6K-chsrWBQKLNlWbZ/inputs/manifest.json
Opus 5.5
256
7
scenes/135yCY9mek1lgvxr6K-chsrWBQKLNlWbZ/metadata/parts.json
scenes/135yCY9mek1lgvxr6K-chsrWBQKLNlWbZ/metadata/articulation.json
scenes/135yCY9mek1lgvxr6K-chsrWBQKLNlWbZ/metadata/measurements.json
scenes/135yCY9mek1lgvxr6K-chsrWBQKLNlWbZ/metadata/physics.json
scenes/135yCY9mek1lgvxr6K-chsrWBQKLNlWbZ/metadata/materials.json
scenes/135yCY9mek1lgvxr6K-chsrWBQKLNlWbZ/simulation/environment.xml
scenes/135yCY9mek1lgvxr6K-chsrWBQKLNlWbZ/simulation/world.sdf
street_corridor
false
scenes/135yCY9mek1lgvxr6K-chsrWBQKLNlWbZ/simulation/environment.usda
Arts venue frontage
18aW1aG0WSpntTSmqsgvL2A6eyZUF7aGF
scenes/18aW1aG0WSpntTSmqsgvL2A6eyZUF7aGF/simulation/environment_pbr.usda
scenes/18aW1aG0WSpntTSmqsgvL2A6eyZUF7aGF/base/model.glb
"\"\"\"Digital twin: sidewalk, '235' courtyard and playground frontage of the Third Street\nMusic Sc(...TRUNCATED)
✓
✓
✓
scenes/18aW1aG0WSpntTSmqsgvL2A6eyZUF7aGF/inputs/manifest.json
Opus 5.5
586
10
scenes/18aW1aG0WSpntTSmqsgvL2A6eyZUF7aGF/metadata/parts.json
scenes/18aW1aG0WSpntTSmqsgvL2A6eyZUF7aGF/metadata/articulation.json
scenes/18aW1aG0WSpntTSmqsgvL2A6eyZUF7aGF/metadata/measurements.json
scenes/18aW1aG0WSpntTSmqsgvL2A6eyZUF7aGF/metadata/physics.json
scenes/18aW1aG0WSpntTSmqsgvL2A6eyZUF7aGF/metadata/materials.json
scenes/18aW1aG0WSpntTSmqsgvL2A6eyZUF7aGF/simulation/environment.xml
scenes/18aW1aG0WSpntTSmqsgvL2A6eyZUF7aGF/simulation/world.sdf
street_corridor
true
scenes/18aW1aG0WSpntTSmqsgvL2A6eyZUF7aGF/simulation/environment.usda
Oculus plaza
18qtBG3qGQUoIkM2LW-bEQR6fwzP2I9cm
scenes/18qtBG3qGQUoIkM2LW-bEQR6fwzP2I9cm/simulation/environment_pbr.usda
scenes/18qtBG3qGQUoIkM2LW-bEQR6fwzP2I9cm/base/model.glb
"\"\" \n\n \n (...TRUNCATED)
✓
✓
✓
scenes/18qtBG3qGQUoIkM2LW-bEQR6fwzP2I9cm/inputs/manifest.json
Opus 5.5
352
4
scenes/18qtBG3qGQUoIkM2LW-bEQR6fwzP2I9cm/metadata/parts.json
scenes/18qtBG3qGQUoIkM2LW-bEQR6fwzP2I9cm/metadata/articulation.json
scenes/18qtBG3qGQUoIkM2LW-bEQR6fwzP2I9cm/metadata/measurements.json
scenes/18qtBG3qGQUoIkM2LW-bEQR6fwzP2I9cm/metadata/physics.json
scenes/18qtBG3qGQUoIkM2LW-bEQR6fwzP2I9cm/metadata/materials.json
scenes/18qtBG3qGQUoIkM2LW-bEQR6fwzP2I9cm/simulation/environment.xml
scenes/18qtBG3qGQUoIkM2LW-bEQR6fwzP2I9cm/simulation/world.sdf
open_public_space
false
scenes/18qtBG3qGQUoIkM2LW-bEQR6fwzP2I9cm/simulation/environment.usda
Shopfront and crosswalk
1EhnbBlIOGf2k2mvSjwZVJOVYC92LqWDO
scenes/1EhnbBlIOGf2k2mvSjwZVJOVYC92LqWDO/simulation/environment_pbr.usda
scenes/1EhnbBlIOGf2k2mvSjwZVJOVYC92LqWDO/base/model.glb
"\"\"\"Digital twin: 4th Avenue x East 12th Street corner, Manhattan (Apple Bank / Crunch).\n\nBRIEF(...TRUNCATED)
✓
✓
✓
scenes/1EhnbBlIOGf2k2mvSjwZVJOVYC92LqWDO/inputs/manifest.json
Opus 5.5
508
5
scenes/1EhnbBlIOGf2k2mvSjwZVJOVYC92LqWDO/metadata/parts.json
scenes/1EhnbBlIOGf2k2mvSjwZVJOVYC92LqWDO/metadata/articulation.json
scenes/1EhnbBlIOGf2k2mvSjwZVJOVYC92LqWDO/metadata/measurements.json
scenes/1EhnbBlIOGf2k2mvSjwZVJOVYC92LqWDO/metadata/physics.json
scenes/1EhnbBlIOGf2k2mvSjwZVJOVYC92LqWDO/metadata/materials.json
scenes/1EhnbBlIOGf2k2mvSjwZVJOVYC92LqWDO/simulation/environment.xml
scenes/1EhnbBlIOGf2k2mvSjwZVJOVYC92LqWDO/simulation/world.sdf
street_corridor
true
scenes/1EhnbBlIOGf2k2mvSjwZVJOVYC92LqWDO/simulation/environment.usda
Residential entrance and steps
1JVjwAoHfLk1fHguJopKYQcI3Rz1PKOvs
scenes/1JVjwAoHfLk1fHguJopKYQcI3Rz1PKOvs/simulation/environment_pbr.usda
scenes/1JVjwAoHfLk1fHguJopKYQcI3Rz1PKOvs/base/model.glb
"\"\"\"Digital twin: one residential street block in Jersey City (capture 1JVjwAoHfLk1fHguJopKYQcI3R(...TRUNCATED)
✓
✓
✓
scenes/1JVjwAoHfLk1fHguJopKYQcI3Rz1PKOvs/inputs/manifest.json
Opus 5.5
896
37
scenes/1JVjwAoHfLk1fHguJopKYQcI3Rz1PKOvs/metadata/parts.json
scenes/1JVjwAoHfLk1fHguJopKYQcI3Rz1PKOvs/metadata/articulation.json
scenes/1JVjwAoHfLk1fHguJopKYQcI3Rz1PKOvs/metadata/measurements.json
scenes/1JVjwAoHfLk1fHguJopKYQcI3Rz1PKOvs/metadata/physics.json
scenes/1JVjwAoHfLk1fHguJopKYQcI3Rz1PKOvs/metadata/materials.json
scenes/1JVjwAoHfLk1fHguJopKYQcI3Rz1PKOvs/simulation/environment.xml
scenes/1JVjwAoHfLk1fHguJopKYQcI3Rz1PKOvs/simulation/world.sdf
street_corridor
true
scenes/1JVjwAoHfLk1fHguJopKYQcI3Rz1PKOvs/simulation/environment.usda
Riverfront promenade
1MGCVydaEAOq0RP7CnMn3E4b24Ait_M5e
scenes/1MGCVydaEAOq0RP7CnMn3E4b24Ait_M5e/simulation/environment_pbr.usda
scenes/1MGCVydaEAOq0RP7CnMn3E4b24Ait_M5e/base/model.glb
"\"\"\"Digital twin: riverside park promenade beside a 3-tier parking garage (capture\n1MGCVydaEAOq0(...TRUNCATED)
✓
✓
✓
scenes/1MGCVydaEAOq0RP7CnMn3E4b24Ait_M5e/inputs/manifest.json
Opus 5.5
369
1
scenes/1MGCVydaEAOq0RP7CnMn3E4b24Ait_M5e/metadata/parts.json
scenes/1MGCVydaEAOq0RP7CnMn3E4b24Ait_M5e/metadata/articulation.json
scenes/1MGCVydaEAOq0RP7CnMn3E4b24Ait_M5e/metadata/measurements.json
scenes/1MGCVydaEAOq0RP7CnMn3E4b24Ait_M5e/metadata/physics.json
scenes/1MGCVydaEAOq0RP7CnMn3E4b24Ait_M5e/metadata/materials.json
scenes/1MGCVydaEAOq0RP7CnMn3E4b24Ait_M5e/simulation/environment.xml
scenes/1MGCVydaEAOq0RP7CnMn3E4b24Ait_M5e/simulation/world.sdf
open_public_space
false
scenes/1MGCVydaEAOq0RP7CnMn3E4b24Ait_M5e/simulation/environment.usda
Scaffolded sidewalk
1P-7EEXp2XCpKqtI3MaLy_t7KEncktjV2
scenes/1P-7EEXp2XCpKqtI3MaLy_t7KEncktjV2/simulation/environment_pbr.usda
scenes/1P-7EEXp2XCpKqtI3MaLy_t7KEncktjV2/base/model.glb
"\"\" \n\n \n (...TRUNCATED)
✓
✓
✓
scenes/1P-7EEXp2XCpKqtI3MaLy_t7KEncktjV2/inputs/manifest.json
Opus 5.5
166
1
scenes/1P-7EEXp2XCpKqtI3MaLy_t7KEncktjV2/metadata/parts.json
scenes/1P-7EEXp2XCpKqtI3MaLy_t7KEncktjV2/metadata/articulation.json
scenes/1P-7EEXp2XCpKqtI3MaLy_t7KEncktjV2/metadata/measurements.json
scenes/1P-7EEXp2XCpKqtI3MaLy_t7KEncktjV2/metadata/physics.json
scenes/1P-7EEXp2XCpKqtI3MaLy_t7KEncktjV2/metadata/materials.json
scenes/1P-7EEXp2XCpKqtI3MaLy_t7KEncktjV2/simulation/environment.xml
scenes/1P-7EEXp2XCpKqtI3MaLy_t7KEncktjV2/simulation/world.sdf
street_corridor
true
scenes/1P-7EEXp2XCpKqtI3MaLy_t7KEncktjV2/simulation/environment.usda
Residential shops and garages
1cPFmcynZnIRhl765fA4CyQzkiAy4rPhc
scenes/1cPFmcynZnIRhl765fA4CyQzkiAy4rPhc/simulation/environment_pbr.usda
scenes/1cPFmcynZnIRhl765fA4CyQzkiAy4rPhc/base/model.glb
"\"\"\"Digital twin: the sidewalks of 1525-1543 Bath Ave, Brooklyn, and the side street behind\nthe (...TRUNCATED)
✓
✓
✓
scenes/1cPFmcynZnIRhl765fA4CyQzkiAy4rPhc/inputs/manifest.json
Opus 5.5
718
27
scenes/1cPFmcynZnIRhl765fA4CyQzkiAy4rPhc/metadata/parts.json
scenes/1cPFmcynZnIRhl765fA4CyQzkiAy4rPhc/metadata/articulation.json
scenes/1cPFmcynZnIRhl765fA4CyQzkiAy4rPhc/metadata/measurements.json
scenes/1cPFmcynZnIRhl765fA4CyQzkiAy4rPhc/metadata/physics.json
scenes/1cPFmcynZnIRhl765fA4CyQzkiAy4rPhc/metadata/materials.json
scenes/1cPFmcynZnIRhl765fA4CyQzkiAy4rPhc/simulation/environment.xml
scenes/1cPFmcynZnIRhl765fA4CyQzkiAy4rPhc/simulation/world.sdf
street_corridor
false
scenes/1cPFmcynZnIRhl765fA4CyQzkiAy4rPhc/simulation/environment.usda
Brick homes and crosswalk
1giQzzRD71Qih-D0Kg-107HJX1KjjKZx5
scenes/1giQzzRD71Qih-D0Kg-107HJX1KjjKZx5/simulation/environment_pbr.usda
scenes/1giQzzRD71Qih-D0Kg-107HJX1KjjKZx5/base/model.glb
"\"\"\"Digital twin: residential street corner, outer-borough New York City.\n\nBRIEF\n R1 Faithfu(...TRUNCATED)
✓
✓
✓
scenes/1giQzzRD71Qih-D0Kg-107HJX1KjjKZx5/inputs/manifest.json
Opus 5.5
603
10
scenes/1giQzzRD71Qih-D0Kg-107HJX1KjjKZx5/metadata/parts.json
scenes/1giQzzRD71Qih-D0Kg-107HJX1KjjKZx5/metadata/articulation.json
scenes/1giQzzRD71Qih-D0Kg-107HJX1KjjKZx5/metadata/measurements.json
scenes/1giQzzRD71Qih-D0Kg-107HJX1KjjKZx5/metadata/physics.json
scenes/1giQzzRD71Qih-D0Kg-107HJX1KjjKZx5/metadata/materials.json
scenes/1giQzzRD71Qih-D0Kg-107HJX1KjjKZx5/simulation/environment.xml
scenes/1giQzzRD71Qih-D0Kg-107HJX1KjjKZx5/simulation/world.sdf
street_corridor
false
scenes/1giQzzRD71Qih-D0Kg-107HJX1KjjKZx5/simulation/environment.usda
Neighborhood park
1hMoHmA-KJVleTI6UFeNI__dS4JvJNls9
scenes/1hMoHmA-KJVleTI6UFeNI__dS4JvJNls9/simulation/environment_pbr.usda
scenes/1hMoHmA-KJVleTI6UFeNI__dS4JvJNls9/base/model.glb
"\"\"\"Digital twin: NYC Parks sitting area in autumn (capture 1hMoHmA-KJVleTI6UFeNI__dS4JvJNls9).\n(...TRUNCATED)
✓
✓
✓
scenes/1hMoHmA-KJVleTI6UFeNI__dS4JvJNls9/inputs/manifest.json
Opus 5.5
320
2
scenes/1hMoHmA-KJVleTI6UFeNI__dS4JvJNls9/metadata/parts.json
scenes/1hMoHmA-KJVleTI6UFeNI__dS4JvJNls9/metadata/articulation.json
scenes/1hMoHmA-KJVleTI6UFeNI__dS4JvJNls9/metadata/measurements.json
scenes/1hMoHmA-KJVleTI6UFeNI__dS4JvJNls9/metadata/physics.json
scenes/1hMoHmA-KJVleTI6UFeNI__dS4JvJNls9/metadata/materials.json
scenes/1hMoHmA-KJVleTI6UFeNI__dS4JvJNls9/simulation/environment.xml
scenes/1hMoHmA-KJVleTI6UFeNI__dS4JvJNls9/simulation/world.sdf
open_public_space
true
scenes/1hMoHmA-KJVleTI6UFeNI__dS4JvJNls9/simulation/environment.usda
End of preview. Expand in Data Studio

Interactive Real-to-Sim Environments for Embodied AI

Explore the scenes in your browser · Download a scene · Cite

If you use this dataset in your research, please cite it (BibTeX below).

The browser viewer covers the original 20 results: rotate the base 3D geometry, inspect semantic annotations and joint definitions, and view static images of the PBR-textured USD scenes. The original 20 results include textured USD files. The 64 results added since then have base appearance only; textured_usd is empty for them. Their previews show the base exported GLB.

Building on Wanderland's real-world urban captures, we release 84 outdoor simulation reconstructions of 42 distinct scenes reconstructed with Nova3D, adding object-level semantics, articulated mechanisms, physical properties, and editable scene programs.

This collection uses Wanderland (Liu et al., CVPR 2026), developed by researchers at NYU’s AI4CE Lab and Cornell University. We thank Xinhao Liu, Jiaqi Li, Prof. Chen Feng, and the Wanderland team for making their captures available.

Nova3D reconstructs these scenes by generating and executing procedural code from Wanderland's released observations. The code defines individual objects and parts, their semantic labels and dimensions, collision geometry, physical properties, and articulated joints. Researchers can edit the code to change scene geometry, layouts, and physical parameters.

The collection includes reconstruction inputs, executable scene source, annotations, PBR assets, and exports for Isaac Sim, MuJoCo, Gazebo, and PyBullet.

The collection covers street and corridor scenes, open public spaces, and other outdoor settings. There are 84 results across 42 distinct capture IDs: every scene has one result made with a mesh input and one made from photos and video alone.

Each scene used eight original photos and a compressed video assembled from all released images for that capture. These are image sequences rather than original video recordings. For 42 results the inputs also included scan-derived geometry converted to GLB; the other 42 used photos and video without a mesh. The supplied inputs and preparation records are included with each scene.

Research status: These are generated reconstructions, not verified ground-truth digital twins. Some geometry, semantic labels, and physical properties are inferred from the available observations. The dataset provides editable scene representations and simulator exports, but geometric fidelity, physical accuracy, and robot-policy performance have not yet been systematically benchmarked.

Wanderland captures beside Nova reconstructions

Wanderland captures beside Nova reconstructions

Contributions

Contribution Included in this release Research Applications
Object and part decomposition Individually identified geometry, object inventory and static/movable declarations. Select, move, replace or remove specific scene components.
Semantic annotations Part names, semantic classes and instance references linked to generated geometry. Associate observations and task targets with identified entities.
Spatial annotations Part bounds, dimensions, positions and scene measurements. Calculate clearance, object size and spatial relationships.
Articulated mechanisms Hinged and sliding parts with pivots, axes, travel limits and damping. Define opening, closing and passage tasks.
Rigid-body and contact physics Movable bodies, collision geometry, mass, inertia, friction and restitution. Simulate contact and motion; vary physical parameters systematically.
Executable scene source Geometry programs and design specifications alongside the assets. Inspect construction logic and create controlled layout, geometry and mechanism variants.
Native simulator exports USD, MuJoCo MJCF, Gazebo SDF, PyBullet files and GLB. Use the reconstruction in different simulator and visualization toolchains.
PBR appearance and enrichment Textured USD, surface materials and additional scene assets. Configure appearance and lighting for visual simulation.
Construction records Generation inputs, source, inventories, physical declarations and file hashes. Inspect provenance, reproduce configurations and diagnose failures.

Dataset Schema

Scene IDs are the original Wanderland capture identifiers. A scene ID can appear twice for mesh/no-mesh counterparts; use mesh_in_input and each row’s file paths to choose the result. Added no-mesh counterparts live under scenes/<scene_id>/variants/no_mesh/; mesh-input results for scenes first released without a mesh live under scenes/<scene_id>/variants/mesh/. Paths are relative to the dataset repository.

Column Meaning
preview_image Rendered overview embedded in the interactive dataset viewer.
environment Human-readable description of the scene.
scene_id Original Wanderland capture ID, preserved exactly. Base and textured variants share it.
textured_usd Path to the PBR USD entrypoint and associated scene enrichment.
base_glb Path to the browser-oriented base GLB.
source_code Full Python scene source. The downloadable file is source/asset.py in the result’s folder.
isaac_sim_export ✓: native Isaac Sim USD files and their resources are included.
mujoco_export ✓: native MuJoCo MJCF files and their resources are included.
gazebo_export ✓: native Gazebo SDF world/model files and their resources are included.
inputs Path to the input file list, hashes and original Wanderland capture references.
generation_model Model used for scene generation: Opus 5.5.
part_count Number of exported base-scene pieces. Several pieces can belong to one object.
joint_count Number of base-scene parts declaring hinge, slide or slider motion. One record per joint-bearing part.
parts Path to full part records, IDs, names and available properties.
articulation Path to joint-bearing part records, including pivots, axes and motion limits.
measurements Path to generated geometry bounds and dimensions, in metres.
physics Path to declared world and body physical parameters.
materials Path to material resources, attribution and format information.
isaac_usd Path to the base Isaac Sim USD export (environment.usda), available for all 84 results. Separate from optional PBR textured_usd. File availability does not certify an Isaac Sim runtime test.
mujoco_mjcf Path to the MuJoCo environment XML.
gazebo_sdf Path to the Gazebo world SDF.
category Broad setting: street/corridor, open public space or other outdoor.
mesh_in_input true when a mesh was supplied as generation input; false otherwise. This records supplied evidence, not proof of successful consumption.

Download

Install the download library:

pip install huggingface_hub

Download one complete scene, including its textures and dependencies:

from huggingface_hub import snapshot_download

scene_id = "135yCY9mek1lgvxr6K-chsrWBQKLNlWbZ"
folder = snapshot_download(
    repo_id="raresense/nova-wanderland",
    repo_type="dataset",
    revision="main",
    allow_patterns=[f"scenes/{scene_id}/*", "LICENSE", "NOTICE", "CREDITS.md"],
    local_dir="robotics-scene-reconstructions",
)

This downloads both results for the scene: one in scenes/<scene_id>/ and the other in scenes/<scene_id>/variants/mesh/ or variants/no_mesh/. For one result only, use that result’s folder (the folder of its inputs path) in allow_patterns. To download all 84 results, omit allow_patterns.

Tool Open inside the result’s folder (scenes/<scene_id>/ or scenes/<scene_id>/variants/<mesh or no_mesh>/)
Isaac Sim, textured simulation/environment_pbr.usda
Isaac Sim, base simulation/environment.usda
MuJoCo simulation/environment.xml
Gazebo simulation/world.sdf
PyBullet simulation/run_pybullet.py
GLB viewer base/model.glb

Keep each scene folder together to preserve relative texture and mesh paths.

Citation

If you use this dataset in your research, please cite it, together with Wanderland, whose captures it builds on:

@misc{nova2026wanderland,
  title        = {Interactive Real-to-Sim Environments for Embodied AI},
  author       = {{Nova3D / RareSense}},
  year         = {2026},
  version      = {1.5.0},
  howpublished = {Hugging Face dataset},
  url          = {https://huggingface.co/datasets/raresense/nova-wanderland},
  note         = {Derived from Wanderland captures; please include the repository revision you used}
}

@inproceedings{liu2026wanderland,
  title     = {Wanderland: Geometrically Grounded Simulation for Open-World Embodied AI},
  author    = {Liu, Xinhao and Li, Jiaqi and Deng, Youming and Chen, Ruxin and Zhang, Yingjia and Ma, Yifei and Guo, Li and Li, Yiming and Zhang, Jing and Feng, Chen},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages     = {1041--1052},
  year      = {2026}
}

The same entries are in CITATION.bib.

Documentation and attribution

Files · Citation · Credits and licenses

Apache-2.0, with included Poly Haven resources under CC0.


Nova3D Website · Open Nova Studio

Downloads last month
6,274

Space using raresense/nova-wanderland 1