The jbDATA dataset is now available!

With the release of the jbDATA dataset, the project consortium is making the final key component of its research available to the public. The dataset was specifically developed to support the validation and analysis of modern perception systems for automated and connected driving. Above all, its focus on rare and safety-critical driving situations sets it apart from other public datasets.

The dataset combines real-world and synthetically generated data in a hybrid approach. The recordings come from various project partners and vehicles and include different sensor configurations with camera, radar, and LiDAR systems. This allows multimodal perception systems to be studied and evaluated under realistic conditions.
Key areas of focus for the long-tail scenarios include unusual objects in road traffic, diverse and adverse weather conditions, and recordings that are underrepresented in conventional datasets. The dataset is supplemented by high-quality ground-truth annotations and pseudolabels based on ASAM OpenLabel.

By publishing this dataset, jbDATA is contributing to the further development of data-driven methods for secure and robust AI applications in the mobility sector. The dataset is now available on the project website: just better Data – Dataset