Three product families. Plus Custom Capture.
PRISM Vision-Language datasets. SABER Vision-Language-Action datasets. Simulation-ready USD environments. Custom Capture engagements. All built on the same dual-stream capture platform.
PRISM: VLM Datasets
10 modalities · richly annotated · dual-stream
PRODUCT FAMILY · PRISM

PRISM — Vision-Language Model Datasets

Richly annotated dual-stream datasets for fine-tuning frontier Vision-Language Models on real-world physical AI tasks.

Annotation Depth: 10 Modalities per Frame
Every PRISM dataset ships with this full annotation stack.
1
Raw Omnidirectional Exocentric Capture
Full-sphere Exocentric RGB + depth from Alia 360° sensor
360° RGB Dense Depth IMU Timestamps
2
Raw Egocentric Capture
RGB + depth from Ego camera(s)
RGB Dense Depth IMU Timestamps
3
3D Reconstruction
Dense neural 3D scene from omnidirectional input
Point Clouds Mesh Gaussian Splats 3D Layout 3D Scene Graphs
4
Spatial Semantics
Navigable paths, obstacles, zones, and surfaces
Path Maps Obstacle Class Floor Plans Zone Labels
5
Object Semantics
Per-object identity, class, pose, and attributes
3D Bounding Boxes Instance Masks SKU Labels 6-DoF Pose
6
Physics Metadata
Object mass, friction, deformability — sim-transfer-ready
Mass Estimates Friction Deformability Collision Mesh
7
Agent Tracking
People, carts, and robots — trajectories over time
Body Keypoints Trajectories Re-ID
8
Skills & Activities
What each agent is doing — pick, place, scan, stack, mop …
500+ Skills Verb–Object Pairs Temporal Spans Role Tags
9
Temporal Context
Time-of-day, traffic density, seasonal and layout variants
Rush / Off-Peak Restocking Layout Changes Lighting
10
Ego-Exo Synchronization
RGB Video Synchronization
RGB Time Sync Ego-Exo Action Mapping
SABER: VLA Datasets
3 action modalities · dual-stream
PRODUCT FAMILY · SABER

SABER — Vision-Language-Action Model Datasets

Dual-stream action datasets for fine-tuning Vision-Language-Action models on real-world physical AI tasks. Built from natural human behavior — not teleoperated demos.

Annotation Depth: 3 Action Modalities per Frame
SABER captures the action-specific data VLA models need — frame-level hand and body pose, contact events, and human-to-robot retargeting. PRISM and SABER are distinct dataset types. Buy them separately, or pair them for full VLM + VLA coverage.
1
Hand Pose & Trajectory
Frame-level manipulation actions from egocentric view
Gripper State Contact Events Hand Pose Force Proxies
2
Body Pose & Trajectory
Full-body pose and movement from egocentric view
Body Keypoints Trajectories Contact Events
3
Human-to-Robot Retargeting
Conversion of human trajectories to standard robot joint data
Robot Motion Robot Control Supports Unitree G1, Fourier and many more
Simulation Assets
Photorealistic USD environments and assets, built from real-world dual-stream capture.

Simulation-ready environments and objects in OpenUSD format, built from DreamVu's real-world dual-stream captures. Photorealistic, physics-equipped, and compatible with any USD simulator.

1
Photorealistic USD environments
Digital twins of any real-world space
2
Individual object assets
Geometry + PBR textures
3
Sim-ready physics
Mass, friction, deformability, collision meshes per object
4
Layout & topology data
Spatial graphs, planograms, navigable paths
5
Lighting & layout variants
Multiple presets per scene
6
Object identity & labels
Class, category, customer-defined attributes per asset
7
OpenUSD-native
Drops into Isaac Sim, IsaacLab, Omniverse, MuJoCo, or any USD-compatible simulator
Custom Capture
On-site data collection & processing

DreamVu deploys to your facility with our Alia 360° capture rigs and full annotation pipeline. You get the same multi-layer annotation stack applied to your specific environment, operations, and use cases.

📷
On-Site Capture
Alia 360° rigs deployed to your location — warehouses, factories, retail, hospitals, or any operational environment.
🎯
Custom Annotation
Full 7- or 9-layer annotation stack tailored to your domain-specific skills, objects, and workflows.
🛠
3D Reconstruction
NuRec pipeline produces dense 3D scenes — point clouds, meshes, and Gaussian splats from your facility.
USD Digital Twin
Your environment converted to simulation-ready OpenUSD assets for any USD-compatible simulator.
📊
Model Fine-Tuning
Optional foundational model fine-tuning on your custom dataset for domain-specific world models or robot policies.
🔒
Exclusive License
Custom capture data is exclusively yours — never shared, sublicensed, or made available to others without your consent.

Ready to build with real-world data?

Talk to our team about datasets, simulation assets, or custom capture for your environment.

Talk to Us