CSG-Stump: A Learning Friendly CSG-Like Representation for Interpretable Shape Parsing
Abstract
Generating an interpretable and compact representation of 3D shapes from point clouds is an important and challenging problem. This paper presents CSG-Stump Net, an unsupervised end-to-end network for learning shapes from point clouds and discovering the underlying constituent modeling primitives and operations as well. At the core is a three-level structure called {\em CSG-Stump}, consisting of a complement layer at the bottom, an intersection layer in the middle, and a union layer at the top. CSG-Stump is proven to be equivalent to CSG in terms of representation, therefore inheriting the interpretable, compact and editable nature of CSG while freeing from CSG's complex tree structures. Particularly, the CSG-Stump has a simple and regular structure, allowing neural networks to give outputs of a constant dimensionality, which makes itself deep-learning friendly. Due to these characteristics of CSG-Stump, CSG-Stump Net achieves superior results compared to previous CSG-based methods and generates much more appealing shapes, as confirmed by extensive experiments. Project page: https://kimren227.github.io/projects/CSGStump/
Cite
@article{arxiv.2108.11305,
title = {CSG-Stump: A Learning Friendly CSG-Like Representation for Interpretable Shape Parsing},
author = {Daxuan Ren and Jianmin Zheng and Jianfei Cai and Jiatong Li and Haiyong Jiang and Zhongang Cai and Junzhe Zhang and Liang Pan and Mingyuan Zhang and Haiyu Zhao and Shuai Yi},
journal= {arXiv preprint arXiv:2108.11305},
year = {2021}
}
Comments
Accepted to ICCV 2021