Boreas Autonomous Driving Dataset
Autonomous Space Robotics Lab, University of Toronto
A year of repeated drives with lidar, radar, and camera in sun, rain, and snow.
Autonomous driving sequences recorded across a year on a repeated route in all weather, with a 128-beam lidar, 5 MP camera, 360-degree radar, and post-processed GNSS ground truth. Aimed at benchmarking long-term localization and odometry across sensors, with object detection planned.
Access
- Cost
- FreeNo charge to download; hosted in an AWS S3 bucket.
- License
- Open licenseCC BY 4.0
- Sign-up
- No sign-up
- How you get it
- API and Bulk downloadsee documentation
What's in it
- One row is
- One driving sequence with synchronized sensor frames
- Covers
- Repeated route in Toronto, Canada
- Years
- about one year
- Updated
- Irregular
- Size
- Not stated
Quality
Rare in covering snow and seasonal change with precise ground truth. Large raw sensor data and specialized formats.
- Large sequences
- Single repeated route
Connects to
Questions it could answer
- How much does localization accuracy degrade in snow?
- Does radar odometry hold up where lidar fails?