lukasvst/dm-vio

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1224 stars · 210 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

DM-VIO is a C++ implementation of Delayed Marginalization Visual-Inertial Odometry, a direct sparse visual-inertial odometry method published in IEEE RA-L 2022. It estimates camera and IMU trajectory from monocular video and IMU data, with a ROS wrapper and live Realsense camera support.

Use cases

• estimate camera trajectory from monocular video and IMU data
• run visual-inertial odometry on EuRoC or TUM-VI datasets
• track a Realsense camera in real time
• process rosbags offline with a ROS wrapper
• reproduce results from the DM-VIO research paper
• build SLAM or drone navigation on top of a VIO backend

When to choose

• you need accurate monocular VIO with IMU, especially with scale drift handling
• you want a research-grade odometry system with dataset evaluation support
• you use Realsense cameras or ROS and want a ready-made wrapper

When to avoid

• you need a full SLAM system with loop closure and mapping
• you want RGB-D or stereo odometry out of the box
• you need a plug-and-play solution without building GTSAM and C++ dependencies from source

Facets

library · maturity maintenance

computer-vision simulation math robotics computer-vision autonomous-vehicles cpp visual-inertial-odometry slam state-estimation gtsam research-code camera-tracking imu linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="lukasvst/dm-vio")