List of Datasets

Other

YCB-Video

Known-object six-degree pose estimation

Cluttered RGB-D object videos with model-assisted pose annotations and YCB object identities.

Metadata: Complete Metadata source documented Released 2017

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

Quick facts

Vehicle type
Other
Environment
Indoor
Data origin
Real-world
Sensor count
2
Ground truth
Available
Calibration
Parameters Only
Annotations
Available

Sensor overview

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

Cameras

2 reported

IMUs

Not 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 Other Manual initialization and model/depth optimization 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.
Hybrid
Reference notes
Object-to-camera 6D pose is initialized manually, refined with models/shared depth and tracked across a static object arrangement. This is not external navigation truth; BOP converts model origins/units and evaluation membership.

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

Author toolbox documents intrinsic_matrix and factor_depth in original meta.mat files. BOP redistribution supplies camera parameters with its own units/conventions; complete raw sessions are not established.

Known or intentional limitations

Reported constraints and characteristics to check before using the dataset.

Ninety-two original real RGB-D videos contain twenty-one object identities. Eighty training videos and twelve held-out videos differ from selected evaluation keyframes and synthetic augmentation; this entry describes the physical-video core.

The old author project URL returns 404. The linked maintained BOP distribution provides reformatted real train/test archives, while its test subset, model origins/units and annotation conventions differ from original YCB-Video.

BOP warns against the original 80k synthetic training images because object poses overlap test poses; later PBR images are a separate augmentation product. Neither adds recorded videos or changes physical camera count.

Object references reuse depth/model assumptions and can have rolling-shutter, alignment, symmetry and occlusion errors. Raw calibration sessions, native depth cadence, device count and independent pose error remain unknown.

Cameras

Camera 1

Model
Asus Xtion Pro Live
Setup
Rgbd
Count
1
Effective cameras
2
Modality
Rgbd
Resolution
640 x 480
Rate
30 Hz
Shutter
Rolling
Lens type
Not Reported
HFOV
Not reported
VFOV
Not reported
Notes
Original 640x480 cropped RGB at 30 fps from an Asus Xtion Pro Live RGB-D acquisition. Depth scale and intrinsics are provided per frame; separate native depth cadence and physical device count remain unverified.

IMUs

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
Original model-assisted 6D object poses, object IDs and label masks; author MAT metadata stores intrinsics, depth scale and camera/object transforms. BOP versions use JSON poses, modal masks and amodal boxes with transformed object models.
Class count
21
Semantic segmentation Not reported
Instance segmentation Available
Object detection Not reported
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
xiang2018posecnn

BibTeX

@inproceedings{xiang2018posecnn,
        Author = {Yu Xiang and Tanner Schmidt and Venkatraman Narayanan and Dieter Fox},
        Title = {{PoseCNN}: A Convolutional Neural Network for {6D} Object Pose Estimation in Cluttered Scenes},
        booktitle = {Robotics: Science and Systems (RSS)},
        Year = {2018}
    }

Dataset corrections

Notice a specific issue after reviewing the page?

Report an inaccuracy