The dataset contains 30 concrete scenarios, 10 per each logical scenario definition.
- Pedestrian Scenario: It involves three actors. A pedestrian unexpectedly crosses the street while the ego-vehicle and another vehicle are driving in their corresponding lanes.
- Bypassing Scenario: The ego vehicle needs to perform an overtake by changing its lane to bypass a slower vehicle.
- Intersection Scenario: The ego vehicle approaches a 4-way intersection when another vehicle performs a left turn in the intersection.
Each concrete scenario has a .json file with the ASAM OpenLABEL defining all the scenario information (actor information, sensor calibration, etc.). In the streams folder of each scenario, the RGB frames taken from a pinhole camera set at the front of the vehicle, and 3D pointclouds from three different LiDAR sensors (16, 32 and 64 channels) mounted at the same location on the roof of the vehicle are stored.
scenario_type/
├─ scenario1 /
├─ .../
├─ scenario10/
│ ├─ scenario10.json
│ ├─ streams/
│ │ ├─ camera_front/
│ │ │ ├─ xxxxxxxxx.png
│ │ ├─ lidar_16/
│ │ │ ├─ xxxxxxxxx.pcd
│ │ ├─ lidar_32/
│ │ │ ├─ xxxxxxxxx.pcd
│ │ ├─ lidar_64/
│ │ │ ├─ xxxxxxxxx.pcd
An example.ipynb notebook is provided with some basic instructions on how to load a given scene. The Projection class can be used to project the LiDAR points and cuboids to the camera image. The code has been tested in conda with Python3.10.
The licensing of this dataset is subject to the Unreal Engine End User License Agreement (EULA). Please refer to the Unreal Engine EULA for specific details regarding the usage and distribution of the dataset. The EULA sets the terms and conditions for using Unreal Engine and any content generated using it, including this dataset.
As a user of this dataset, it is your responsibility to comply with the terms specified in the Unreal Engine EULA. Ensure that you read and understand the EULA before using the dataset for any purpose.
This dataset was generated employing the CARLA Simulator, on Unreal Engine 4. The dataset was generated as part of the work presented in the paper "Synthetic Dataset Generation Using Logical Scenario Files for Automotive Perception Testing". The paper is to be published in the 36th IEEE Intelligent Vehicles Symposium (IEEE IV 2025).
If you use this dataset in your research, publications, or academic work, please cite the following paper:
Proceedings yet to be published
For any questions or inquiries, please contact [email protected]
To download the dataset, please fill the following formulary: Download Form