The performance seems to be pretty poor on any section of the data, but I am particularly interested in the offroad section (roughly 1250 seconds into Route B), where tracking is lost very quickly every time and the pointcloud is pretty indecipherable.
I was wondering if, in your opinions, this might be something rectifiable by the correct settings, or if this environment is simply outside of the scope of good performance for LSD-SLAM.
Hi,
First off, thanks for putting together (and just as importantly, maintaining) such a high quality SLAM system!
I have been attempting to use LSD-SLAM with SFU's Mountain Dataset, which can be found here: https://autonomylab.github.io/sfu-mountain-dataset/
The performance seems to be pretty poor on any section of the data, but I am particularly interested in the offroad section (roughly 1250 seconds into Route B), where tracking is lost very quickly every time and the pointcloud is pretty indecipherable.
I was wondering if, in your opinions, this might be something rectifiable by the correct settings, or if this environment is simply outside of the scope of good performance for LSD-SLAM.
Thanks for your time! James