All datasets

Behavioral Dataset: Multimodal Video, Depth & Actions

Behavioral Dataset: Multimodal Video, Depth & Actions

About

Human-teleoperated demonstrations of everyday household tasks for embodied AI and robotic manipulation workflows. Each episode includes synchronized RGB video, depth, and instance segmentation.

Specifications

File types
mp4
Video format
H.264
Frame rate
30fps
Resolution
720x720 sample media
Media metadata
Resolution, frame rate, codec, duration, bitrate, and file size included in samples
Depth
8-bit (quantized)
Segmentation
Instance ID maps
Multi view
Head & wrist cameras
Simulator
OmniGibson
Format
LeRobot

Sample details

Washing a baseball cap

A teleoperated household-task clip with first-person view, RGB video, depth, and segmentation.

Camera tripod setup

A teleoperated task clip where a camera is attached to a tripod.

Slicing vegetables

A kitchen-task clip covering food handling and cutting motions.

Poseidon AI, Inc. © 2026
All rights reserved