An open-source visual-inertial odometry system that estimates camera motion and sparse 3D maps from camera and IMU data.
XIVO is an open-source visual-inertial odometry system that estimates the 6-degree-of-freedom pose of a camera and builds sparse 3D maps from video and IMU data. It solves the problem of real-time motion tracking and environment mapping for robotic and augmented reality applications by fusing visual features with inertial measurements. The system includes auto-calibration of camera-IMU alignment and runs efficiently at high frame rates.
Researchers, students, and developers working on robotics, autonomous systems, augmented reality, or computer vision who need a real-time visual-inertial odometry solution. It's particularly suited for educational purposes due to its simplified, pedagogical design.
Developers choose XIVO for its high performance (140 FPS), real-time capabilities, and robust auto-calibration features. As an open-source, simplified version of the Corvis system, it provides a clear implementation for learning and prototyping visual-inertial odometry without sacrificing efficiency.
X Inertial-aided Visual Odometry
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Achieves 140 FPS on stored data and 1-7 ms latency on live streams, as stated in the README, enabling efficient real-time processing for robotics and AR.
Automatically calibrates camera-IMU relative pose and time-stamp alignment, reducing manual setup effort and improving robustness in sensor fusion.
Simplified from the Corvis system for educational purposes, with a clear implementation that helps students and researchers learn visual-inertial odometry principles.
Most dependencies, except OpenCV, are included in the thirdparty directory, streamlining the build process and reducing external setup complexity.
Lacks loop-closure and global re-localization features that are part of the full Corvis system, limiting its use for large-scale or long-term mapping tasks.
Primarily built and tested on Ubuntu 20.04 with g++9, as noted in the README, making it less portable and potentially challenging for cross-platform development.
The software is provided free only for research with no warranties; commercial use requires contacting UCLA TDG, adding legal and procedural hurdles.