Scale

Canadian Adverse Driving Conditions Dataset

Open-source dataset for autonomous driving in wintry weather.

Download Dataset →
Learn More About Scale →
  • Scene #1
  • Scene #2
  • Scene #3
  • Scene #4
  • Scene #5
  • Scene #6
Select Scene

Overview

High quality data for adverse driving conditions

a Waterloo car one driving on the snow
a Waterloo car two driving on the snow

Data Collection

Complex Driving Scenarios in Adverse Conditions

Car Setup

Vehicle, Sensor and Camera Details

    • 10 Hz capture frequency
    • 1/1.8” CMOS sensor of 1280x1024 resolution
    • Images are stored as PNG
    8
    Wide Angle Cameras
    • 10 Hz capture frequency
    • 32 channels
    • 200m range
    • 360° horizontal FOV; 40° vertical FOV (-25° to +15°)
    1
    LiDAR
  • 1
    Post-processed GPS and IMU

More on Autonomoose: The University of Waterloo's self-driving research platform.

Sensor Calibration

Data alignment between sensors and cameras

  • LiDAR extrinsics

  • Camera extrinsics

  • Camera intrinsic calibration

  • IMU extrinsics

Data Annotation

Complex Label Taxonomy

Scene #1

The resulting accuracy is consistently higher than what a human or synthetic labeling approach can achieve independently as measured against seven rigorous quality areas for each annotation.

The CADC includes 3D Bounding boxes for X object classes and a rich set of class attributes related to X, Y. For detailed definitions of each class and example images, please see the annotation instructions.

Instances Per Label

28194

Cars

62851

Pedestrians

20441

Trucks

4867

Bus

4808

Garbage Containers on Wheels

3205

Traffic Guidance Objects

705

Bicycle

638

Pedestrian With Object

75

Horse and Buggy

26

Animals

Get Started with CADC Dataset

View our paper here and download the development kit.

If you use our dataset please cite our paper.

Download Dataset