Datasets / Physical AI

Paired Egocentric RGB and IMU Dataset

20,000 hours · synchronized video and motion

Learn from visual context and participant motion recorded together through calibrated egocentric cameras and synchronized inertial sensors.

Dataset specification

  • 2016×1512 RGB video at 30 fps
  • Dual 6-axis IMU streams at 800 Hz
  • Calibrated camera intrinsics and extrinsics
  • Nanosecond timestamps on a common clock
  • VRS, MP4, and Parquet or CSV sensor exports

Intended uses

  • Embodied visual and inertial representation learning
  • Motion-aware perception for robotics
  • Activity and task-phase understanding

Scope notes

  • This dataset contains participant motion and visual observations; robot action commands are provided through the separate teleoperation program.
  • The activity distribution and selected delivery subset are confirmed during review.

Collection scope

Activities

  • Real-world activities across varied environments
  • Task, activity, action, phase, and completion labels

Capture

  • 2016×1512 RGB at 30 fps
  • Dual 6-axis IMU streams at 800 Hz

Catalog volume

  • 20,000 hours
  • Activity mix and delivery subset confirmed during review

Alignment and delivery

  • Calibrated intrinsics and extrinsics with nanosecond common-clock timestamps
  • VRS, MP4, and Parquet or CSV sensor exports

Annotation & metadata fields

  • Task, action, and activity
  • Task phase and completion
  • Participant and environment labels
  • Sensor-quality labels

Capture methodology

  • Consented participants wear egocentric cameras and IMU sensors during real-world activities
  • RGB video and inertial records share a common timestamp reference
  • Camera intrinsics and extrinsics accompany the sensor records

Provenance & rights chain

  • Recorded by consented participants or operators
  • Licensing and the selected delivery scope are reviewed with Datoric

Quality, duplicates & PII

How submissions are reviewed and cleaned before they are accepted into the dataset.

  • Dropped-frame checks
  • Sensor-dropout checks
  • Video and inertial stream alignment

Formats & delivery

  • VRS recordings
  • MP4 video
  • Parquet or CSV sensor exports
  • Camera calibration and timestamp metadata

Rights & license scope

Licensed by Datoric for robotics training, with final scope governed by the agreement for the selected delivery.

Frequently asked

What are the video and inertial rates?
The catalog specifies 2016×1512 RGB video at 30 fps and dual 6-axis IMU streams at 800 Hz.
How are the streams aligned?
The records use nanosecond common-clock timestamps with calibrated camera intrinsics and extrinsics.
Which delivery formats are supported?
The catalog lists VRS recordings plus MP4 video and Parquet or CSV sensor exports. The selected package is confirmed during review.

Related dataset specifications