QuadricsNet: Learning Concise Representation for Geometric Primitives in Point Clouds

TL;DR

QuadricsNet introduces an end-to-end learning framework that efficiently represents 3D point cloud primitives using quadrics, enabling robust and concise geometric primitive parsing with a new comprehensive dataset.

Contribution

It is the first to propose a learning-based approach for quadric-based primitive representation in point clouds, integrating geometric attributes for effective supervision.

Findings

Effective primitive representation with only 10 parameters

Robustness demonstrated on a new comprehensive dataset

Outperforms existing methods in primitive parsing accuracy

Abstract

This paper presents a novel framework to learn a concise geometric primitive representation for 3D point clouds. Different from representing each type of primitive individually, we focus on the challenging problem of how to achieve a concise and uniform representation robustly. We employ quadrics to represent diverse primitives with only 10 parameters and propose the first end-to-end learning-based framework, namely QuadricsNet, to parse quadrics in point clouds. The relationships between quadrics mathematical formulation and geometric attributes, including the type, scale and pose, are insightfully integrated for effective supervision of QuaidricsNet. Besides, a novel pattern-comprehensive dataset with quadrics segments and objects is collected for training and evaluation. Experiments demonstrate the effectiveness of our concise representation and the robustness of QuadricsNet. Our code is available at \url{https://github.com/MichaelWu99-lab/QuadricsNet}