gaps when used on large-scale point cloud

[dxbbwan]

Thanks for your excellent work!

When using test_large.py, the generated point cloud has obvious gaps, which may cause by clustering. I wonder is there a good way to eliminate this, do you have any advice? thx

[iegorval]

Were you able to apply this method to large point clouds?
Due to KMeans used, it seems to me as a no-go to even try on some real-world dataset where it is usually rather 50 millions than thousands points.