CLARU
haosulab2023apache-2.0

ManiSkill Data

A unified benchmark for learning generalizable robotic manipulation skills with 20 task families, 2000+ diverse object models, and 4M+ demonstration frames powered by SAPIEN.

Downloads97
Likes2

Why This Matters for Physical AI

ManiSkill provides a large-scale diverse manipulation benchmark with high-speed simulation and demonstrations to enable research in vision-based reinforcement learning, imitation learning, and generalizable robotic manipulation skills.

Technical Profile

Modalities
rgbdepth
Environment
simulation
Task Types
manipulationgraspingpick_and_place
License
apache-2.0
Part of the ManiSkill family

Access

View on HuggingFaceRead Paper

Need custom rgb data?

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

Request a Sample Pack

Related Datasets

ManiSkill Demonstrations

A collection of demonstration datasets for ManiSkill environments, including trajectories from motion planning, reinforcement learning, and teleoperation sources. The dataset contains simulation-based robotic manipulation tasks with multiple control modes.

rgbproprioception
2K downloadsapache-2.0

ManiSkill-HAB TidyHouse Dataset

A dataset of robot manipulation trajectories for pick and place tasks in simulated home environments, generated from reinforcement learning policies with filtering. Contains 18K episodes with 3.6M transitions across multiple object categories.

rgbdepthproprioception
479 downloadsmit

ManiSkill-HAB PrepareGroceries Dataset

A dataset of 18K episodes with 3.6M transitions for robotic manipulation tasks including picking and placing grocery items in home environments using whole-body control.

rgbdepthproprioception
264 downloadsmit

AI2THOR Scene/Object Dataset for ManiSkill

A modified version of the Habitat Synthetic Scenes Dataset containing 10,000+ scenes and 10,000+ objects optimized for use with the ManiSkill robotic simulation framework.

rgbdepth
92 downloadscc-by-4.0

Xperience-10M

A large-scale egocentric multimodal dataset of human experience containing 10 million interactions and 10,000 hours of synchronized first-person recordings with six video streams, audio, stereo depth, camera pose, hand mocap, full-body mocap, IMU, and hierarchical language annotations for embodied AI, robotics, and world modeling research.

rgbaudiodepthproprioception+3
2.3M downloadsother

Open-H-Embodiment

A multi-embodiment community-driven dataset of paired kinematics and video for training AI autonomy models in surgical robotics and ultrasound applications. The dataset includes tabletop exercises, clinical procedures, and simulations of healthcare robotics tasks.

rgbproprioception
118K downloads2026cc-by-4.0