UCLA-VAIL2025apache-2.0
NavBench-GS: 3D Gaussian Splatting Scenarios for Visual Navigation
A collection of real-world 3D Gaussian Splatting (3DGS) scenarios for benchmarking visual navigation models, containing reconstructed outdoor scenes with USD meshes, 3DGS point clouds, and navigation waypoints.
Downloads12
Why This Matters for Physical AI
NavBench-GS provides real-world 3D scene reconstructions and navigation benchmarks that enable training and evaluation of visual navigation models, bridging perception and embodied action in outdoor environments.
Technical Profile
- Modalities
- point_cloud3d_mesh
- Robot Embodiments
- mobile_robot
- Action Space
- waypoint_navigation
- Environment
- outdoor
- Task Types
- navigation
- Data Format
- USD
- Annotation Types
- navigation_waypoints
- License
- apache-2.0
Access
Need custom point_cloud data?
Claru builds purpose-built datasets for outdoor applications with dense human annotations and quality assurance.
Request a Sample Pack