Aurora Multi-Sensor Dataset: Urban Driving Data from Pittsburgh

Between January 2017 and February 2018, Aurora (via Uber ATG) and the University of Toronto captured this large-scale multi-sensor dataset in the Pittsburgh, PA metropolitan area. It includes data from a 64-beam LiDAR sensor and seven cameras, with ground truth for localization. The dataset spans all four seasons and various weather, time-of-day, and traffic conditions.

Use Cases

• Develop autonomous vehicle localization systems based on multi-sensor data and accurate ground truth.
• Train 3D scene reconstruction models based on LiDAR point clouds and 360-degree camera imagery.
• Construct and compress high-definition (HD) maps based on the detailed urban environment scans.
• Evaluate perception algorithms across diverse conditions based on metadata for weather, time, and traffic.

Strengths

• Captured over a 13-month period spanning all four seasons, providing temporal diversity.
• Includes data from a 64-beam LiDAR and seven cameras, including a stereo pair and 360-degree coverage.
• Contains highly accurate localization ground truth and rich metadata such as weather and semantic segmentation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and total dataset size are unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

Aurora (via Uber ATG) in collaboration with the University of Toronto.

Collection Method

Captured by a sensor-equipped vehicle in an urban environment.

Time Range

January 2017 to February 2018

Geography

Metropolitan area of Pittsburgh, Pennsylvania, USA