NeurIPS 2020

UCSG-NET- Unsupervised Discovering of Constructive Solid Geometry Tree


Meta Review

All reviewers are enthusiastic about the idea of the paper. One reviewer is concerned about sufficient empirical validation, but states that it is a matter of where is the threshold for sufficient empirical validation. AC finds the paper sufficiently novel and interesting, and agrees with the reviewers that it points to a promising future direction. Thus AC recommends acceptance. In the camera ready version, in addition to revising the paper based on the rebuttal, the authors should pay particular attention to strengthening empirical validation. In particular, the authors should include more results to show stats on the types of trees that it learns and provide a better description on how well the method works.