Closed AlexReimann closed 1 year ago
Basically a fix for #20, as this has a much smaller traversability.
Thanks for your PR! What was the issue you tried to fix with this filter?
With your original implementation, the border of the invalid cells have large step value. By checking the validity of the neighboring cells, the map becomes like this. Which one do you intend?
Which one do you intend?
The second one, but I wonder why I didn't see this before.
We definitely have "NAN" cells from clearing, which should currently cause an issue then. Can you supply a patch?
Okay, please have a look at this branch, which is a modified one from your PR. https://github.com/leggedrobotics/elevation_mapping_cupy/tree/step_filter
Closing we are running now with heavy modifications, so not putting effort into getting this back into upstream anymore.
Adds a
StepFilter
, which simply results in a layer which has the highest differences to the neighboring cells.Mostly copy & paste from the
MinFilter
.Has settings for the neighbor cell radius and if negative differences should be included