A toolbox for calibrating multi-camera, visual-inertial, and rolling shutter sensor systems in robotics.
Kalibr is a toolbox for calibrating multi-sensor systems, specifically addressing multi-camera, visual-inertial, and rolling shutter camera calibration. It solves spatial, temporal, and intrinsic calibration problems to ensure accurate sensor fusion, which is essential for robotics applications like SLAM and autonomous navigation.
Robotics researchers, engineers, and developers working with visual-inertial systems, autonomous vehicles, or drones who need precise sensor calibration for reliable state estimation and perception.
Developers choose Kalibr for its comprehensive, research-backed calibration methods, support for a wide range of camera models, and ability to handle complex multi-sensor setups with rolling shutter cameras, all available as open-source with Docker and ROS integration.
The Kalibr visual-inertial calibration toolbox
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Solves multi-camera, visual-inertial, and multi-IMU calibration in a unified toolbox, as detailed in the README's list of addressed problems.
Supports a wide range of camera models, including double sphere models, which were added via community contributions and are highlighted in the news section.
Based on peer-reviewed papers from conferences like ICRA and CVPR, ensuring methods are validated and academically sound, as referenced in the README.
Specifically addresses full intrinsic calibration for rolling shutter cameras, a feature added via PR and supported by cited research.
Requires setting up via Docker or a ROS workspace, which the wiki guides but can be time-consuming and prone to dependency issues, especially with older Ubuntu versions.
Tightly integrated with ROS 1 and Docker, making it unsuitable for projects outside these ecosystems or with strict deployment constraints.
Uses continuous-time batch estimation methods, which are not designed for real-time calibration and require pre-recorded data, limiting dynamic applications.