Voxel512026cc-by-4.0

DreamTac (FiftyOne multimodal MCAP)

A Franka Emika Panda dataset of 703 contact-rich tabletop manipulation trajectories with synchronized multi-view RGB cameras and vision-based tactile sensors on gripper fingertips. Contains 506,078 frames across 12 task types with language instructions.

Downloads882
Episodes703
Hours7.03
Likes1

Why This Matters for Physical AI

This dataset provides high-quality contact-rich manipulation data with vision-based tactile sensing, enabling robots to learn from multimodal sensory feedback for dexterous object manipulation tasks.

Technical Profile

Modalities
rgbtactileproprioceptionlanguage
Robot Embodiments
Franka Panda
Environment
lab
Task Types
manipulationpick_and_placegraspingpushinginsertioncleaningstackingcutting
Episodes
703
Total Hours
7.03
Data Format
MCAP
Annotation Types
language_instructionsaction_labels
License
cc-by-4.0
Part of the DreamTac 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