The USyd Campus Dataset
Since March 2018, we started collecting driving data on a weekly basis over the University of Sydney (USyd) campus and surroundings. This USyd Campus Dataset contains more than 60 weeks drives and is continuously updated.
It includes multiple sensor modalities (camera, lidar, GPS, IMU, wheel encoder, steering angle, etc.) and covers various environmental conditions as well as diverse changes to illumination, scene structure, and pedestrian/vehicle traffic volumes.
A comprehensive set of tools for using the dataset is available:
Dataset highlight video:
We have summarized each dataset information in the spreadsheet. You can use it as a reference to find the data for your need.
We also provided a set of development toolkets to facilitate the visualization and utilization of our dataset. Please refer to the following section for more details.
The tools have been combined into a single metapackage available at the following address:
Existing toolkit links
Tool set version 2
A new set of tools is under active development to overcome some of the shortcomings of the first version.
- Written in c++ to massively increase speed, and reduce memory consumption
- Generic toolset – loads all camera topics, and can operate in different resolutions by subscribing to the appropriate topics
- Motion correction for lidar
- Projection of lidar to camera and visa versa
ROS playback of bags + video files: https://gitlab.acfr.usyd.edu.au/its/dataset_tools.git
Image stitching node: https://gitlab.acfr.usyd.edu.au/nvidia/panorama_generator.git
Lidar motion correction: https://gitlab.acfr.usyd.edu.au/its/lidar_motion_correction.git
Lidar/camera projection: https://gitlab.acfr.usyd.edu.au/its/lidar_camera_projection.git
please see the individual repos for instructions.
We have one customized frame_info topic which will be used for all camera related tools. Please download this first and build it in your catkin workspace.
The USyd Campus Dataset can be used for design, train, validate and test different algorithms. In this section, we provide some possible applications using this dataset.