UCSG-Net -- Unsupervised Discovering of Constructive Solid Geometry Tree

TL;DR

UCSG-Net is an unsupervised model that automatically discovers constructive solid geometry trees from 3D shapes, enabling high-fidelity reconstruction of complex, non-convex objects without prior parse tree supervision.

Contribution

The paper introduces UCSG-Net, a novel unsupervised approach for extracting CSG trees from 3D shapes, eliminating the need for supervised parse tree annotations during training.

Findings

Achieves high-quality 3D shape reconstruction.

Produces interpretable CSG parse trees.

Works effectively on 2D and 3D autoencoding tasks.

Abstract

Signed distance field (SDF) is a prominent implicit representation of 3D meshes. Methods that are based on such representation achieved state-of-the-art 3D shape reconstruction quality. However, these methods struggle to reconstruct non-convex shapes. One remedy is to incorporate a constructive solid geometry framework (CSG) that represents a shape as a decomposition into primitives. It allows to embody a 3D shape of high complexity and non-convexity with a simple tree representation of Boolean operations. Nevertheless, existing approaches are supervised and require the entire CSG parse tree that is given upfront during the training process. On the contrary, we propose a model that extracts a CSG parse tree without any supervision - UCSG-Net. Our model predicts parameters of primitives and binarizes their SDF representation through differentiable indicator function. It is achieved jointly with discovering the structure of a Boolean operators tree. The model selects dynamically which operator combination over primitives leads to the reconstruction of high fidelity. We evaluate our method on 2D and 3D autoencoding tasks. We show that the predicted parse tree representation is interpretable and can be used in CAD software.