List of Datasets

Automotive

Boreas

Multi-season odometry and metric localization

Repeated autonomous-driving routes across a year, with camera, lidar, scanning radar, processed IMU, pose references and a sunny-weather 3D-box subset.

Metadata: Complete Metadata curator reviewed

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

Supplementary

Not reported

Quick facts

Vehicle type
Automotive
Environment
Outdoor
Data origin
Real-world
Sensor count
4
Ground truth
Available
Calibration
Parameters Only
Annotations
Available

Sensor overview

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

Cameras

1 reported

IMUs

1 reported

GNSS

Not reported

LiDAR

1 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 GNSS / INS Applanix POSPac with RTX corrections · Applanix POS LV 220 Not reported 200 Hz Partial · Public Training-Sequence Reference Poses; Test Reference Poses Are Withheld. 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
Accuracy and notes
Accuracy notes
POSPac estimated residual RMS: typically below 5 cm in nominal conditions, but up to 20-40 cm in urban canyons; satellite dependent. Table 2 nominal device specification is 2-4 cm RMS. These are not uniform independent reference-accuracy bounds.
Reference notes
One POSPac/RTX trajectory product distributed at 200 Hz in Applanix frame and at camera/lidar/radar frame times; do not count each output cadence as a separate reference. Wheel encoder and GNSS are reference-generation equipment unless independent raw release is established.

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

camera0_intrinsics.yaml and rectified P_camera.txt; homogeneous intersensor transforms in calib/. Camera intrinsics and camera/lidar extrinsics use MATLAB tools; radar rotation uses stationary Fourier-Mellin matches and translation uses CAD; Applanix/lidar alignment uses batch fitting. A separate raw calibration package is not established.

Known or intentional limitations

Reported constraints and characteristics to check before using the dataset.

Sensor acquisition is asynchronous despite shared UTC clocks.

Lidar motion distortion and radar motion/Doppler distortion require compensation.

Radar firmware changes range and range-bin spacing across sequences.

Pose residual estimates depend on satellite visibility and worsen in urban canyons.

3D boxes cover a sunny-weather subset; test labels are withheld and interpolated frames have lower fidelity.

Camera images are rectified and anonymized by default.

Cameras

Camera 1

Model
FLIR Blackfly S BFS-U3-51S5C
Setup
Mono
Count
1
Effective cameras
1
Modality
Rgb
Resolution
2448 x 2048
Rate
10 Hz
Shutter
Not Reported
Lens type
Not Reported
HFOV
81 deg
VFOV
71 deg
Notes
Rectified PNG images; default face/license-plate blurring. Shutter type is not established by the paper.

IMUs

Applanix POS LV 220 post-processed IMU

Model
Not reported
Count
1
Accel / gyro
200 / 200 Hz

GNSS

Not reported

LiDAR

Velodyne Alpha-Prime

Model or name
Velodyne Alpha-Prime
Count
1
Dimensions
3
Rate
10 Hz
Channels
128
Range
300 m

Additional sensors

Radar

Model
Navtech CIR304-H
Count
1
Rate
4 Hz
Notes
360-degree scanning radar; polar PNG with per-azimuth times and encoder values. Firmware-dependent nominal range/resolution: newer 250 m / 0.0438 m; older 200 m / 0.0596 m. Motion and Doppler distortion persist.

Annotations

Reported annotation availability and task support.

Availability
Available
Format
Per-frame text rows: uuid, type, dx, dy, dz, x, y, z, yaw; lidar-frame 3D boxes with within-scene track UUIDs. Boreas-Objects-V1 only: 7111 annotated frames at 5 Hz / 326180 boxes / 53 scenes (37 training, 16 test); test labels withheld; sunny-weather subset.
Class count
[Car Cyclist Pedestrian Misc]
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
burnett_ijrr23

BibTeX

@article{burnett_ijrr23,
  title={Boreas: A Multi-Season Autonomous Driving Dataset},
  author={Burnett, Keenan and Yoon, David J. and Wu, Yuchen and Li, Andrew Zou and Zhang, Haowei and Lu, Shichen and Qian, Jingxing and Tseng, Wei-Kang and Lambert, Andrew and Leung, Keith Y. K. and Schoellig, Angela P. and Barfoot, Timothy D.},
  journal={The International Journal of Robotics Research},
  volume={42}, number={1--2}, pages={33--42}, year={2023},
  doi={10.1177/02783649231160195}
}

Dataset corrections

Notice a specific issue after reviewing the page?

Report an inaccuracy