Visual (Inertial) Odometry in ROS noetic
Hello everyone
I am looking for V(I)O algorithms that are available for ROS Noetic. The VO is to be performed with a monocular camera and an IMU. I have already found and tried the following algorithms:
- SVO2.0: https://github.com/srv/viso2
- MSCKF: https://github.com/daniilidis-group/m...
- OKVIS: https://github.com/ethz-asl/okvis_ros
- VINS-Fusion: https://github.com/HKUST-Aerial-Robot...
- Viso2: https://github.com/srv/viso2
So far I got Viso2 and MSCKF working, however results aren't great. Any ideas for more up-to-date V(I)O would be appreciated!
PS. I am aware some SLAM algorithms also incorporate V(I)O (like RTAB-map and ORB-SLAM2). What I'm looking for is an algorithm that only does V(I)O. This because I want to compare different available algorithms.
Thanks in advance
Hendrik
What do you mean by standalone implementation?
I changed the post to hopefully make it a bit more clear. Basically, I want to compare some V(I)O-techniques with eachother. If it is possible to only use the V(I)O part of ORB-SLAM, than I will take a look at it as well.
You can use only the visual data for ORB slam, but it wont exploit the IMU information. To combine both source of information, did you think of writing a pose graph with edges being created from the IMU data and transformations you calculate from the visual data? Would that work for you?
This is an interesting solution, I'll certainly give it a try. Do you have any knowledge about additional algorithms that I could use? Only having one algorithm in a comparison is not ideal :).