The paper introduced a new encoding of point clouds based on graph neural networks. It received scores of 8,6,8. The reviews for this paper were overly positive, and do not sufficiently convey the strengths of the paper to match the rating. The AC read the paper, rebuttal, and all reviews. The novelty of this paper is limited, particularly in light of missed relevant related work: Qi et al, 3D Graph Neural Networks for RGBD Semantic Segmentation, ICCV'19 This paper needs to be cited and discussed. Second, the authors need to improve the writing, in particular, make the writing slightly more scientific by omitting words such as "nowadays", and overly using the word "one". The AC will not overturn the reviewers' recommendation, however, urges the authors to polish their camera ready, and add discussion to relevant work.