SweepNet: Unsupervised Learning Shape Abstraction via Neural Sweepers

TL;DR

SweepNet introduces an unsupervised neural approach for shape abstraction using sweep surfaces, combining a novel parameterization with a differentiable neural sweeper to effectively capture and simplify complex geometric structures.

Contribution

The paper presents a new unsupervised neural method for shape abstraction that employs a compact sweep surface parameterization and a differentiable neural sweeper architecture.

Findings

Outperforms existing methods in shape abstraction quality.

Requires only 14 float numbers for shape representation.

Demonstrates effective shape editing and preservation of details.

Abstract

Shape abstraction is an important task for simplifying complex geometric structures while retaining essential features. Sweep surfaces, commonly found in human-made objects, aid in this process by effectively capturing and representing object geometry, thereby facilitating abstraction. In this paper, we introduce \papername, a novel approach to shape abstraction through sweep surfaces. We propose an effective parameterization for sweep surfaces, utilizing superellipses for profile representation and B-spline curves for the axis. This compact representation, requiring as few as 14 float numbers, facilitates intuitive and interactive editing while preserving shape details effectively. Additionally, by introducing a differentiable neural sweeper and an encoder-decoder architecture, we demonstrate the ability to predict sweep surface representations without supervision. We show the superiority of our model through several quantitative and qualitative experiments throughout the paper. Our code is available at https://mingrui-zhao.github.io/SweepNet/