DM-VIO: Delayed Marginalization Visual-Inertial Odometry [Code online]

Published at IEEE Robotics and Automation Letters.
Project Page: vision.in.tum.d...
Code online at: github.com/luk...
Authors:
Lukas von Stumberg
Daniel Cremers
Abstract:
We present DM-VIO, a monocular visual-inertial odometry system based on two novel techniques called delayed marginalization and pose graph bundle adjustment. DM-VIO performs photometric bundle adjustment with a dynamic weight for visual residuals. We adopt marginalization, which is a popular strategy to keep the update time constrained, but it cannot easily be reversed, and linearization points of connected variables have to be fixed. To overcome this we propose delayed marginal- ization: The idea is to maintain a second factor graph, where marginalization is delayed. This allows us to later readvance this delayed graph, yielding an updated marginalization prior with new and consistent linearization points. In addition, delayed marginalization enables us to inject IMU information into already marginalized states. This is the foundation of the proposed pose graph bundle adjustment, which we use for IMU initialization. In contrast to prior works on IMU initialization, it is able to capture the full photometric uncertainty, improving the scale estimation. In order to cope with initially unobservable scale, we continue to optimize scale and gravity direction in the main system after IMU initialization is complete. We evaluate our system on the EuRoC, TUM-VI, and 4Seasons datasets, which comprise flying drone, large-scale handheld, and automotive scenarios. Thanks to the proposed IMU initialization, our system exceeds the state of the art in visual-inertial odometry, even outperforming stereo- inertial methods while using only a single camera and IMU. The code will be published at vision.in.tum.de/dm-vio

[@airreyz2617]

Nice, looking forward to trying on my devices

[@francoisdev]

Impressive!

[@xudongcao6711]

Wow！ Impressive!

[@zhenglin1872]

Great work! Can it work with equirectangular cameras such as Ricoh Theta V?

[@lichaoxu710]

nice work! could you share the information about the imu used in the tests?

[@cvprtum]

Thanks! You can find details about the used IMUs in the papers of the respective datasets. The EuRoC dataset uses an ADIS16448, the TUM-VI dataset a Bosch BMI160 and the 4Seasons dataset uses an ADIS16465.

[@eddygonzalez6513]

Is there a version for ROS2 ?

[@lidarslam2765]

Awesome......

[@albertomartin6218]

Awesome work! do you think this could work with a GoPro Hero 9 camera? comes with an IMU Bosch BMI260.

[@cvprtum]

The IMU should be sufficient, but the camera probably has a rolling shutter which will decrease the accuracy. With a good camera calibration it might still be worth trying it out though.

[@albertomartin6218]

​@@cvprtum Thank you very much for your answer and for the code, can´t wait to try!

[@engineerasifali]

Amazing

[@hanin2199]

nice!!!!!

[@AndersonSilva-dg4mg]

Interesting. Waiting for the code..

[@cvprtum]

The code is now published.

[@AndersonSilva-dg4mg]

@@cvprtum Hooray, thank you!

[@mtbrust]

flat earth