List of Datasets

Other

Aria Digital Twin

Egocentric 3D perception and real/synthetic digital twins

Egocentric Aria RGB/inertial recordings paired with synthetic digital-twin views, tracked device/object poses, depth, instance labels and human-interaction metadata.

Metadata: Complete Metadata source documented Released 2023

Primary places to inspect, download, read about, or reproduce the dataset.

Quick facts

Vehicle type
Other
Environment
Indoor
Data origin
Hybrid
Sensor count
5
Ground truth
Available
Calibration
Parameters Only
Annotations
Available

Sensor overview

High-level sensor availability before the detailed sensor records below.

Cameras

3 reported

IMUs

2 reported

GNSS

Not reported

LiDAR

Not reported

Radar

Not reported

Magnetometers

Not reported

Barometers

Not reported

UWB

Not reported

Ground truth

Reference scope, method, rate, coverage, and provenance.

Scope Method Reference system Count Rate Coverage Reference provenance
External
Measured using equipment independent of the platform’s normal onboard sensor pipeline.
Derived
Computed primarily from the dataset’s own recorded sensors.
Hybrid
Combines independent reference equipment with onboard measurements.
6DoF Motion capture Aria rigid-body tracking · Optitrack Not reported Not reported Unknown Reference provenance
External
Measured using equipment independent of the platform’s normal onboard sensor pipeline.
Derived
Computed primarily from the dataset’s own recorded sensors.
Hybrid
Combines independent reference equipment with onboard measurements.
External
Reference notes
Calibrated motion-capture device trajectory; object-system validation does not assert device-only trajectory accuracy.
6DoF Motion capture Dynamic object rigid-body tracking · Optitrack Not reported Not reported Unknown Reference provenance
External
Measured using equipment independent of the platform’s normal onboard sensor pipeline.
Derived
Computed primarily from the dataset’s own recorded sensors.
Hybrid
Combines independent reference equipment with onboard measurements.
External
Accuracy notes
Reported mean object-system error 6.83 mm / 1.29 degrees on sampled validation, not a global bound.
Other Other Digital-twin rendered metric depth Not reported Not reported Unknown Reference provenance
External
Measured using equipment independent of the platform’s normal onboard sensor pipeline.
Derived
Computed primarily from the dataset’s own recorded sensors.
Hybrid
Combines independent reference equipment with onboard measurements.
Hybrid
Reference notes
Digitized geometry plus tracked/calibrated viewpoint; synthetic product paired with real images.

Calibration and synchronization

Reported calibration level, reproducibility signals, and supporting notes.

Level
Parameters Only
Processed parameters
Available
Raw calibration data
Not reported
Calibration targets
Not reported

Parameters only

Processed calibration parameters are reported, but raw calibration data is not reported.

Notes

Device camera/inertial calibration and tracked rigid-body transforms are documented. Complete raw sessions and independently validated final residuals remain unestablished.

Known or intentional limitations

Reported constraints and characteristics to check before using the dataset.

Physical RGB/SLAM-camera and inertial measurements are paired with synthetic sensor views from digital twins. Rendered depth/instance products depend on geometry and tracked viewpoints; physical camera counts exclude renderings and external motion-capture equipment.

The original paper describes 150 apartment and 50 office sequences, while maintained packaging splits multi-person recordings into per-device sequences. Current aggregate recording count remains unknown rather than mixing those units.

Object totals conflict across paper sections. Raw eye-camera delivery, native IMU rates and full independent device/object/depth error coverage remain unverified; gaze outputs alone do not establish raw eye imagery.

Motion-capture coverage and human-skeleton products vary by recording. Access uses a signup/links file or dataset explorer, and sampled object-system validation is not a corpus-wide reference-accuracy guarantee.

Cameras

Camera 1

Model
Not reported
Setup
Mono
Count
1
Effective cameras
1
Modality
Rgb
Resolution
1408 x 1408
Rate
30 Hz
Shutter
Not Reported
Lens type
Not Reported
HFOV
Not reported
VFOV
Not reported
Notes
One physical Aria RGB stream at 30 fps. synthetic_video.vrs is its rendered counterpart, not another physical camera.

Camera 2

Model
Not reported
Setup
Stereo
Count
1
Effective cameras
2
Modality
Grayscale
Resolution
Not reported
Rate
30 Hz
Shutter
Not Reported
Lens type
Not Reported
HFOV
Not reported
VFOV
Not reported
Notes
Two physical grayscale SLAM streams at 30 fps; exact image geometry remains unverified. Paired synthetic views do not add physical devices.

IMUs

Aria integrated IMUs

Model
Not reported
Count
2
Accel / gyro
Not reported

GNSS

Not reported

LiDAR

Not reported

Radar

Not reported

Additional sensors

Not reported

Annotations

Reported annotation availability and task support.

Availability
Available
Format
Rendered per-pixel instance IDs, derived 2D boxes, 3D object poses/boxes, human joints/raw markers and gaze vectors
Class count
Not reported
Semantic segmentation Not reported
Instance segmentation Available
Object detection Available
Optical flow Not reported
Depth ground truth Available

Citation

Citation key and BibTeX kept at the end of the page for reference.

Citation key
pan2023ariadigitaltwin

BibTeX

@misc{pan2023ariadigitaltwin,
  author = {{Xiaqing Pan} and {Nicholas Charron} and {Yongqian Yang} and {Scott Peters} and {Thomas Whelan} and {Chen Kong} and {Omkar Parkhi} and {Richard Newcombe} and {Carl Yuheng Ren}},
  title = {Aria Digital Twin: A New Benchmark Dataset for Egocentric 3D Machine Perception},
  year = {2023},
  url = {https://arxiv.org/abs/2306.06362v2}
}

Dataset corrections

Notice a specific issue after reviewing the page?

Report an inaccuracy