Pointnet2_PyTorch
Pointnet2_PyTorch copied to clipboard
[BUG] Error Sampling Result for Furthest Point Sampling
I build the pointnet2_ops on Windows 10, and the version of Python is 3.7. When I use the furthest_point_sample
and gather_operation
to sample points from a dense point cloud, I find that there is a bug in the algorithm:
Here is the dense point cloud:

After use fps to sample 2048 points, there is a hole on the plane:

It happened many times when apply to different models.