I used the tutorial_euroc on the master branch of maplab to process the EUROC dataset in stereo mode.
And lastly I exported the vimap into plain csv files including vertices.csv.
All resulting trajectories in vertices.csv look good except V2_03_difficult.
I recall that some paper mentioned that some algorithm cannot process this sequence due to a few missing frames, eg., VIO aided SfM. But in rviz, the ROVIO provides continuous dense odometry estimates.
So I guess the missing poses in vertices.csv or vi map has to do with probably the featureTracker module.
The vimap trajectory in vertices.csv are drawn in the below two figures.
I ran ROVIO twice, and it gave very similar results on V2_03 difficult, so this is not a random phenomenon.
Can you please confirm this or give a fix?
FYI, my OS ubuntu 20.04, ROS Noetic.
The commands to produce above results:
Dear authors,
I used the tutorial_euroc on the master branch of maplab to process the EUROC dataset in stereo mode. And lastly I exported the vimap into plain csv files including vertices.csv. All resulting trajectories in vertices.csv look good except V2_03_difficult. I recall that some paper mentioned that some algorithm cannot process this sequence due to a few missing frames, eg., VIO aided SfM. But in rviz, the ROVIO provides continuous dense odometry estimates. So I guess the missing poses in vertices.csv or vi map has to do with probably the featureTracker module.
The vimap trajectory in vertices.csv are drawn in the below two figures.
![xyz](https://github.com/ethz-asl/maplab/assets/8062561/91ff8a7d-ae3e-468f-9cd8-354778d9cd21)
I ran ROVIO twice, and it gave very similar results on V2_03 difficult, so this is not a random phenomenon. Can you please confirm this or give a fix?
FYI, my OS ubuntu 20.04, ROS Noetic. The commands to produce above results:
The content of exportscript
Regards,