List of Datasets

Other

DexYCB

Hand and object pose estimation

Synchronized multiview RGB-D recordings of tabletop object grasping by ten subjects, with fitted hand and object pose annotations.

Metadata: Complete Metadata needs review Released 2021

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

Download

Not reported

Quick facts

Vehicle type
Other
Environment
Indoor
Data origin
Real-world
Sensor count
16
Ground truth
Available
Calibration
Not reported
Annotations
Available

Sensor overview

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

Cameras

16 reported

IMUs

Not reported

GNSS

Not reported

LiDAR

Not reported

Radar

Not reported

Pitot tube

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 Other Not reported Not reported Not reported Full 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.
Derived
Reference notes
Rigid-object poses fitted using multiview depth, human 2D keypoints and known mesh models; PoseCNN initialization is manually selected. Includes tabletop objects.
Other Other Not reported Not reported Not reported Partial · Annotated Hands 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.
Derived
Accuracy notes
Mean fingertip reprojection errors are below five pixels against human 2D annotations; this is not a metric 3D accuracy guarantee.
Reference notes
Articulated MANO hand pose and 21 joint positions, with subject-specific shape fixed during fitting; 508,384 hand annotations across 581,968 images.

Calibration and synchronization

Reported calibration level, reproducibility signals, and supporting notes.

Level
Not reported
Processed parameters
Not reported
Raw calibration data
Not reported
Calibration targets
Not reported

Calibration data

Calibration reproducibility details are not reported.

Notes

Camera extrinsics are calibrated; delivery of parameters or raw calibration captures is not established by the paper.

Known or intentional limitations

Reported constraints and characteristics to check before using the dataset.

Controlled laboratory capture with constant background; RGB models may generalize poorly to in-the-wild images.

Fast motion and ambiguous human labels can produce large keypoint reprojection errors.

Cameras

capture-rgbd

Model
RealSense D415
Setup
Rgbd
Count
8
Effective cameras
16
Modality
Rgbd
Resolution
640 x 480
Rate
30 Hz
Shutter
Not Reported
Lens type
Not Reported
HFOV
Not reported
VFOV
Not reported
Notes
Both color and depth are recorded at 640 by 480, with all eight views recorded together.

IMUs

Not reported

GNSS

Not reported

LiDAR

Not reported

Radar

Not reported

Pitot tube

Not reported

Additional sensors

Not reported

Annotations

Reported annotation availability and task support.

Availability
Available
Format
Not reported
Class count
21
Semantic segmentation Not reported
Instance segmentation Available
Object detection Available
Optical flow Not reported
Depth ground truth Not reported

Citation

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

Citation key
chao2021dexycb

BibTeX

Not reported

Dataset corrections

Notice a specific issue after reviewing the page?

Report an inaccuracy