List of Datasets

UGV

JackRabbot Dataset and Benchmark (JRDB)

2D and 3D person detection and tracking

Egocentric multimodal robot recordings in indoor and outdoor campus human environments, with pedestrian boxes and time-consistent tracks.

Metadata: Complete Metadata needs review

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

Code

Not reported

Supplementary

Not reported

Quick facts

Vehicle type
UGV
Environment
Indoor, Outdoor
Data origin
Real-world
Sensor count
12
Ground truth
Available
Calibration
Parameters Only
Annotations
Available

Sensor overview

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

Cameras

5 reported

IMUs

1 reported

GNSS

Not reported

LiDAR

4 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.
3D position + heading Manual annotation Not reported Not reported 15 Hz Partial · Public Annotations for Train-Validation Sequences in This Manuscript Version 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
Human 3D boxes on merged Velodyne measurements are annotated at 7.5 Hz and linearly interpolated to 15 Hz; 2D boxes are projected, manually refined and supplemented. The same pipeline supplies corresponding per-sequence 2D/3D track IDs. This is object ground truth, not robot-pose ground truth.

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

The appendix directory tree includes cameras.yaml and defaults.yaml under calibration.

Known or intentional limitations

Reported constraints and characteristics to check before using the dataset.

occlusion

dynamic_objects

Cameras

occam-raw

Model
Not reported
Setup
Stereo
Count
1
Effective cameras
2
Modality
Rgb
Resolution
752 x 480
Rate
15 Hz
Shutter
Not Reported
Lens type
Not Reported
HFOV
Not reported
VFOV
Not reported
Notes
Occam cylindrical stereo suite has two rows of five raw cameras. Each row forms a derived 360-degree panorama; only the upper row and its panorama are annotated. One physical stereo suite is counted; derived stitched products add no sensor systems.

ricoh-theta

Model
Ricoh Theta
Setup
Fisheye
Count
1
Effective cameras
1
Modality
Rgb
Resolution
1280 x 720
Rate
15 Hz
Shutter
Not Reported
Lens type
Not Reported
HFOV
Not reported
VFOV
Not reported
Notes
The source describes a 360-degree fisheye RGB camera without establishing a Theta generation.

zed-rgbd

Model
ZED
Setup
Rgbd
Count
1
Effective cameras
2
Modality
Rgbd
Resolution
672 x 376
Rate
Not reported
Shutter
Not Reported
Lens type
Not Reported
HFOV
Not reported
VFOV
Not reported
Notes
Sensor setup reports 15 fps, whereas captured-data section reports 30 fps. Rate remains unresolved.

IMUs

JackRabbot IMU

Model
Not reported
Count
1
Accel / gyro
Not reported

GNSS

Not reported

LiDAR

Velodyne 16 Puck LITE

Model or name
Velodyne 16 Puck LITE
Count
2
Dimensions
3
Rate
Not reported
Channels
16
Range
Not reported

SICK LMS500-20000

Model or name
SICK LMS500-20000
Count
2
Dimensions
2
Rate
25 Hz
Channels
1
Range
Not reported

Radar

Not reported

Pitot tube

Not reported

Additional sensors

Wheel encoder

Name
Active wheel position encoders
Count
2
Notes
Encoder values from the robot wheels are included; these are not an external reference trajectory.

Other sensor

Name
Arm and head position encoders
Count
1
Notes
Position readings cover arm joints and fingers and the head pan-tilt motors; no physical encoder total is proposed.

Other sensor

Name
Digital microphone
Count
1
Notes
Audio is included among the sensor signals in the dataset.

Annotations

Reported annotation availability and task support.

Availability
Available
Format
json
Class count
1
Semantic segmentation Not reported
Instance segmentation Not reported
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
martinmartin2021jrdb

BibTeX

@misc{martinmartin2021jrdb,
  title={JRDB: A Dataset and Benchmark of Egocentric Robot Visual Perception of Humans in Built Environments},
  author={Roberto Mart\'in-Mart\'in and Mihir Patel and Hamid Rezatofighi and Abhijeet Shenoi and JunYoung Gwak and Eric Frankel and Amir Sadeghian and Silvio Savarese},
  year={2021},
  eprint={1910.11792},
  archivePrefix={arXiv},
  primaryClass={cs.CV},
  note={Version 4, 24 April 2021},
  url={https://arxiv.org/abs/1910.11792v4}
}

Dataset corrections

Notice a specific issue after reviewing the page?

Report an inaccuracy