Trying to understand data pre-processing/neighborhood computation

[mackenzie-warren]

In Section 4 of the paper (https://arxiv.org/abs/1904.02375), there is a description of the neighborhood computation of {q}, by scoring points to ensure regions of the point clouds are sampled uniformly. However, I cannot find the corresponding part of the code. For example, looking at lines 112-114 in semantic3d_seg.py, it looks like the points are just sampled randomly without any weighting. Could you provide some clarification?