Brno Urban Dataset
A multi-sensor dataset for autonomous vehicle and robot navigation, featuring synchronized camera, LiDAR, IMU, and GNSS data collected in urban environments.
What is Brno Urban Dataset?
Brno Urban Dataset is a publicly available multi-sensor dataset collected in urban environments for autonomous vehicle and robot research. It provides synchronized data from cameras, LiDAR, IMU, GNSS, and other sensors to support development in navigation, localization, and perception. The dataset includes diverse weather, time-of-day, and environmental conditions to enable robust algorithm testing.
Researchers, startups, and academic groups working on autonomous driving, mobile robotics, computer vision, and sensor fusion who need real-world data without investing in expensive sensor hardware.
It offers a comprehensive, well-documented, and freely accessible alternative to building custom data acquisition systems, with synchronized multi-sensor data across varied real-world conditions that are rare in other public datasets.
Overview
Navigation and localisation dataset for self driving cars and autonomous robots
Use Cases
Best For
- Training and validating perception models for self-driving cars
- Developing sensor fusion algorithms for localization and mapping
- Benchmarking SLAM (Simultaneous Localization and Mapping) systems
- Researching autonomous navigation in diverse urban environments
- Testing robustness of algorithms across different weather and lighting conditions
- Educational projects in robotics and autonomous systems
Not Ideal For
- Projects requiring millimeter-accurate sensor calibration for precise 3D reconstruction
- Research teams needing extensive, manually verified object annotations for training detection models
- Applications dependent on real-time data streaming or frequently updated datasets
- Environments with strict IT policies that block torrent-based downloads
Pros & Cons
Pros
Includes RGB, IR, multiple LiDARs, IMU, GNSS, and radar, all synchronized with timestamps, enabling comprehensive sensor fusion studies as detailed in the data description table.
Tagged by weather, daytime, and environment types, providing over 80 recordings across sunny, rainy, city, and highway conditions for robust algorithm testing, as shown in the tags distribution table.
Available via torrent at no cost, lowering barriers for startups and universities without budget for expensive hardware, as emphasized in the project philosophy.
The winter extension adds Livox LiDAR and FMCW radar, plus pre-computed YOLO detections, offering specialized data for cold-weather research, noted in the winter extension section.
Cons
Calibration is performed manually as a best guess, with plans for future improvement, which may introduce errors in precise localization tasks, as admitted in the calibrations section.
Issues like misordered IR frames, inverted IMU acceleration sign, and missing LiDAR rows require additional preprocessing and correction, documented in the Known Bugs section.
Distribution via torrent files can be less user-friendly than direct downloads, potentially complicating access for researchers unfamiliar with torrent clients, as indicated in the download instructions.
Object detections are only provided via YOLO in the winter extension, lacking detailed annotations for other recordings or broader object classes, limiting out-of-the-box use for perception tasks.
Frequently Asked Questions
Related Projects
GitHub repositoryAutonomous Vehicle Seasonal Dataset
GitHub repositoryDevkit for the Boreas autonomous driving dataset.
Ford DatasetThe dataset is time-stamped and contains raw data from all the sensors, calibration values, pose trajectory, ground truth pose, and 3D maps. The data is Robot Operating System (ROS) compatible
Oxford RobotCarThe Oxford RobotCar Dataset contains over 100 repetitions of a consistent route through Oxford, UK, captured over a period of over a year
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.