CSGNet: Neural Shape Parser for Constructive Solid Geometry

TL;DR

This paper introduces CSGNet, a neural network that efficiently converts 2D or 3D shapes into constructive solid geometry programs, improving speed, interpretability, and detection accuracy over existing methods.

Contribution

It presents a top-down recurrent neural network approach for shape parsing into CSG programs, enabling training without explicit program annotations.

Findings

Faster shape parsing compared to bottom-up methods.

Produces more interpretable modeling instructions.

Outperforms existing shape detection techniques.

Abstract

We present a neural architecture that takes as input a 2D or 3D shape and outputs a program that generates the shape. The instructions in our program are based on constructive solid geometry principles, i.e., a set of boolean operations on shape primitives defined recursively. Bottom-up techniques for this shape parsing task rely on primitive detection and are inherently slow since the search space over possible primitive combinations is large. In contrast, our model uses a recurrent neural network that parses the input shape in a top-down manner, which is significantly faster and yields a compact and easy-to-interpret sequence of modeling instructions. Our model is also more effective as a shape detector compared to existing state-of-the-art detection techniques. We finally demonstrate that our network can be trained on novel datasets without ground-truth program annotations through policy gradient techniques.