Giters

Pointcept / Pointcept

Pointcept: a codebase for point cloud perception research. Latest works: PTv3 (CVPR'24 Oral), PPT (CVPR'24), OA-CNNs (CVPR'24), MSC (CVPR'23)

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

How to remove point cloud clipping

Ale0311 opened this issue · comments

commented

Hello,

I tried to train the model on my data, following the config from semantic Kitti. The problem is that the model does not seem to be able to detect the far away points to classify them correctly.

I thought it was because of the PointClip class and I removed it from the transforms array. However, the results are exactly the same. I also tried changing the values, so that the point cloud is clipped further away.

My question is: do you cip the point cloud in another place as well? Or what can I do in order to get predictions for the far away objects as well?

Thank you!

commented

In my experience, PointClip is not needed for PTv2 and PTv3 but is necessary for SpUNet. You mentioned that the network can not predict well on far points as predicting far points is a challenge for outdoor perception as usually they are more sparse and sometime not labelled correctly.