jbDATAset
Authors/Creators
Description
jbDATAset
This repository contains ASAM OpenLABEL annotation files for the jbDATAset, a dataset that was generated as part of the just better DATA (jbDATA) research project. The OpenLABEL files provide a standardized interface for the usage of the dataset that comprises raw data from three different vehicles. The annotations provide standardized labels for sequences of the jbDATAset and are intended to facilitate interoperability, reproducibility, and the use of common annotation formats in automated driving research.
___________________________________________________________________________________________________________________
Note
This repository contains OpenLABEL files only. The corresponding raw sensor data (camera, LiDAR, radar, etc.)
must be downloaded separately. Visit the official jbDATA website for more information.
___________________________________________________________________________________________________________________
About the Project
just better DATA (jbDATA) is a publicly funded German research project that develops AI-based methods and tools for efficient, high-quality data recording and generation for autonomous driving. A major result of the project is the jbDATAset. For more information check out the official project website.
About the Dataset
The OpenLABEL annotations provided in this repository correspond to the jbDATAset, a multimodal dataset designed for the development of automated driving functions in challenging and safety-critical driving scenarios.
The dataset contains real world data that was recorded by vehicles from the partners AVL, FZI, and b-plus. The raw data needs to be downloaded separately. Follow the instructions on the website for that: https://www.justbetterdata.de/en/dataset-en/
Repository Structure
The repository is structured as follows:
jbDATAset
├── openlabel
│ ├── AVL/
│ │ ├── sequence_00001.json # any name that corresponds to the sequence
│ │ ├── sequence_00002.json
│ │ └── ...
│ ├── b-plus/
│ └── fzi/
├── AVL # the raw data folder for the subset
├── b-plus # the raw data folder for the subset
└── fzi # the raw data folder for the subset
Each JSON file follows the ASAM OpenLABEL specification and contains annotations for a single sequence. The Openlabel files then reference the corresponding raw data. Make sure to check the names of the raw data folders when you download the raw data. The paths in the OpenLabel files referencing the raw data are given from the jbDATAset base folder.
ASAM OpenLABEL
The annotations are provided in the ASAM OpenLABEL format, an open and vendor-independent standard for describing annotations of automotive and robotics datasets.
OpenLABEL enables standardized representations of:
- Objects
- Coordinate systems
- Annotations
- Paths to raw data
- Temporal information
- making annotations interoperable across tools and datasets.
The specification is available at: https://www.asam.net/standards/detail/openlabel/
Citation
If you use these annotations in your research, please cite:
@inproceedings{jbDATAset,
title={{jbDATA - A Collaborative Framework for Automatic Curation and Creation of a Dataset for AI Functions in Autonomous Driving}},
author={Nowzad, Azarm and Sch{\"o}tz, Michael and K{\"o}nigshof, Hendrik and Aufschl{\"a}ger, Robert and Bally, Fabian and Bogdoll, Daniel and Heinrich, Marc and Huang, Wencan and Karunakaran, Dinish and Kraft, Erwin and Polley, Rupert and Reis, Philipp and R{\"o}sch, Kevin and Schlauch, Christian and Sch{\"o}ll, Martina Katharina and Sch{\"o}rner, Philip and Sigl, Christina and Stiehm, Torsten and Szabados, Peter and Weber, Stefan and Yao, Yue and Zajusch, Tobias and Klein, Nadja},
booktitle={European conference on computer vision (ECCV)},
year={2026},
organization={Springer}
}
License
The OpenLABEL files for the jbDATAset are released under CC-BY SA 4.0 license.
The contextual metadata contains information derived from third-party data sources:
- OpenStreetMap-derived metadata is subject to the Open Data Commons Open Database License 1.0 and the applicable attribution requirements.
OpenStreetMap
A part of the road network and environmental attributes were obtained from OpenStreetMap data. OpenStreetMap data is ©OpenStreetMap contributors and is available under the Open Database License (ODbL) 1.0.
Acknowledgements
These annotation files were generated within the publicly funded just better DATA (jbDATA) project.
The authors gratefully acknowledge all project partners for collecting, processing, and providing the underlying multimodal dataset.
The research leading to these results is funded by the German Federal Ministry for Economic Affairs and Energy within the project “just better DATA Effiziente und hochgenaue Datenerzeugung für KI-Anwendungen im Bereich autonomes Fahren”. The authors would like to thank the consortium for the successful cooperation.