How did you calculate shape/localization error?
Closed this issue · 4 comments
This is not an error.
But I wanted to know how did you calculate localization and shape error. Is there a ground truth for that?
Did you have access to Quadric SLAM code for comparison?
Hi, did you mean the ground truth of the objects?
We annotated the objects manually using the point cloud to get the ground truth. The tool is like this one (https://github.com/XunshanMan/semantic-data-association-tool). However it's hard to use currently, we are planning to make it more user-friendly, and also release further experiment datasets after our paper is accepted.
For QuadricSLAM, its code has not been released publicly, so we reproduced by ourselves according to the paper.
I hope it helps.
@XunshanMan thank you very much. It certainly helped.
Is it possible that you share the QuadricSLAM code? If it's not possible I totally understand. I wanted to check if I can work with some custom datasets.
Thanks for the great work anyways.
@joshi-bharat I think it'll take some time because of the recent events and influenced schedule. I'll share it on the github after I clean it up. :)
Thank you. I am closing this issue now.