mzxuan

Real-World Data — Camera-3 Annotations

Perception annotations for 800 real robot demonstration episodes across four manipulation tasks, including instance masks, object identities, 2D/3D bounding boxes, and role labels for camera 3 only.

Downloads0
Episodes800

Why This Matters for Physical AI

This densely annotated real-world manipulation dataset enables training of perception models for robotic object detection, instance segmentation, and 3D spatial understanding in cluttered manipulation scenarios.

Technical Profile

Modalities
rgbdepthproprioception
Environment
lab
Task Types
manipulationpick_and_place
Episodes
800
Data Format
HDF5
Annotation Types
instance_segmentationobject_detectionsegmentation3d_bounding_boxes2d_bounding_boxeslanguage_instructions
Part of the Real-World Data family

Access

Need custom rgb data?

Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.

Request a Sample Pack

Related Datasets