## A2D2: Audi Autonomous Driving Dataset

An open multi-sensor dataset for autonomous driving research. This
dataset comprises annotated semantic segmentation images and point
clouds, and 3D bounding boxes. We also include three unannotated
sequences with 360° time-synchronized and registered camera and lidar
data, together with corresponding bus data. We hope this
commercial-friendly dataset will further facilitate research and
development in AI and robotics for autonomous driving.


## What is this for?

Our intention is to support academic institutions and commercial
start-ups working on autonomous driving research. Equipping a vehicle,
recording data, and annotating it, is both labour intensive and costly.
As such, this initial phase presents a substantial barrier to entry. By
removing this barrier we would like to enable more institutions to
conduct research in this field.


## Vehicle sensor suite

We have equipped an Audi Q7 e-tron with six cameras and five LiDAR
sensors. This results in 360° camera and LiDAR coverage. Additionally,
we recorded several bus data signals from the vehicle such as velocity,
steering wheel angle, odometry, etc.


## How is the dataset organized

We release A2D2 in three data subsets. Each subset has been tailored to
emphasize one of the following topics:

- Semantic Segmentation
- 3D Object Detection
- Sensor Fusion: Camera, Lidar and Bus

Each subset is described in more detail in its corresponding README file.


## License

A2D2 is released under the CC BY ND 4.0 license. For more information,
please see https://creativecommons.org/licenses/by-nd/4.0/


## Contact

Feel free to contact us at aevdrivingdataset@audi.de


## How can I contribute?

You are welcome to contribute to this dataset in any way that may be
useful to the community. This may range from sharing research results
obtained from using this dataset to fixing annotations in it.