Quantitative metric for tuning

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Quantitative metric for tuning

jkarimi91
What is a quantitative metric rtabmap is optimizing? How does one extract/see that value for a map? I would like to extract that value and use it as a measure for map quality as I tune. This will allow me to tweak parameters and quantitatively see there effect on the map.
Reply | Threaded
Open this post in threaded view
|

Re: Quantitative metric for tuning

matlabbe
Administrator
Hi,

If you are able to setup a ground truth (e.g., with VICON or Optitrack), this data can be fed to rtabmap (by setting ground_truth_frame_id and ground_truth_base_frame_id parameters, which are the origin of the localization system and a fake base frame representing same frame than base_link, but in a different TF tree than the standard map->odom->base_link tree) to compute some quantitative results (like the absolute trajectory error of TUM's RGB-D dataset).

EDIT see this example: https://github.com/introlab/rtabmap_ros/blob/master/launch/tests/rgbdslam_datasets.launch

cheers,
Mathieu