An open-source Python library for automated, safety-informed characterization of trajectory datasets, designed to categorize, score, and mine driving scenarios across heterogeneous trajectory datasets.
The tool provides a dataset-agnostic API built on Pydantic-validated schemas and Hydra-based hierarchical configuration, enabling seamless integration with diverse data sources and experimental pipelines.
Beyond academic prototyping, the library has been deployed in industry settings for real-world scenario mining, where it powers pipelines that ingest, convert, and characterize vehicle event logs while supporting agent criticality analysis through dedicated visualization tools. Distributed as a versioned PyPI package with pre-commit hooks and GitHub Actions CI workflows, it is engineered to meet the code quality and reproducibility standards required for production use.
