You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In my application, the measured points are not equally spaced. Some regions are empty and then there are some regions with rather high density. Is pointnet able to deal with this?
Here is an example image for which I would like to do semantic segmentation:
The text was updated successfully, but these errors were encountered:
Is it robust by design to varying densities or does the performance of the network decrease if the density is not homogeneous? Your studies in the paper seems to be done on samples, where the density of the points is rather high. In my case the interesting regions are sometimes not very dense. Do you expect problems at any stage and how could one heal this?
Hi @Merlin1896 the robustness if by design (of the max pooling layer). It's very insensitive to the density and can deal with varying density very well.
In PointNet++ we need multi-scale layers for varying density cases.
In my application, the measured points are not equally spaced. Some regions are empty and then there are some regions with rather high density. Is pointnet able to deal with this?
Here is an example image for which I would like to do semantic segmentation:
The text was updated successfully, but these errors were encountered: