Learning Manifold Patch-Based Representations of Man-Made Shapes

TL;DR

This paper introduces a novel shape representation for man-made 3D models using deformable parametric patches, enabling effective sketch-based modeling and training with synthetic data.

Contribution

It proposes a new manifold patch-based shape representation compatible with CAD pipelines and deep learning, along with a self-supervised training method to handle artifacts and data scarcity.

Findings

Successfully infers 3D shapes from raster images.

Demonstrates shape interpolation capabilities.

Outperforms related methods in sketch-based modeling tasks.

Abstract

Choosing the right representation for geometry is crucial for making 3D models compatible with existing applications. Focusing on piecewise-smooth man-made shapes, we propose a new representation that is usable in conventional CAD modeling pipelines and can also be learned by deep neural networks. We demonstrate its benefits by applying it to the task of sketch-based modeling. Given a raster image, our system infers a set of parametric surfaces that realize the input in 3D. To capture piecewise smooth geometry, we learn a special shape representation: a deformable parametric template composed of Coons patches. Naively training such a system, however, is hampered by non-manifold artifacts in the parametric shapes and by a lack of data. To address this, we introduce loss functions that bias the network to output non-self-intersecting shapes and implement them as part of a fully self-supervised system, automatically generating both shape templates and synthetic training data. We develop a testbed for sketch-based modeling, demonstrate shape interpolation, and provide comparison to related work.