ScenarioMax is an extension to ScenarioNet that transforms converted data from ScenarioNet into a standardized TFRecord format. It currently supports real-world datasets like Waymo, nuPlan, and nuScenes through a unified interface, making the data compatible with both Waymax, V-Max and GPUDrive.
- Unified Data Interface: Access multiple autonomous driving datasets through a single API
- Data Transformation Pipeline: Convert raw scenario data into a standardized format thanks to ScenarioNet.
- TFRecord Conversion: Generate TFRecord files for efficient processing in Waymax and V-Max. Extended support for the GPUDrive format.
- SDC Path Support: Add and evaluate multiple potential trajectories for self-driving vehicles
- Extensible Architecture: Easily add support for new datasets
| Dataset | Version | Link | Status |
|---|---|---|---|
| Waymo Open Motion Dataset | v1.3.0 | Site | ✅ Full Support |
| nuPlan | v1.1 | Site | ✅ Full Support |
| nuScenes | v1.0 | Site | ✅ Full Support |
| Argoverse | v2.0 | Site | 🚧 WIP |
For dataset setup, you can see the complete ScenarioNet documention.
Support 3.10
pip install -e .pip install -e devkit/nuplan-devkit
pip install -r devkit/nuplan-devkit/requirements.txtpython scenariomax/convert_dataset.py --waymo_src /data/datasets/womd/training/ --dst /data/test --log-level INFO --shard 10 --num-workers 10 --format tfexampleScenarioMax provides a streamlined workflow for:
- Loading raw scenario data from various sources
- Processing the data into a unified format
- Transforming scenarios with additional features
- Exporting to TFRecord format for use in simulation environments
- Initial implementation with support for Waymo, nuPlan, and nuScenes
- TFRecord conversion pipeline for efficient data handling
- SDC paths support
This project is licensed under the MIT License - see the LICENSE file for details.
For issues or feature requests, please open an issue on the repository or contact the project maintainers.