A Systematic Survey on Deep Learning Architectures for Point Cloud Classification and Segmentation

TL;DR

This paper systematically reviews deep learning architectures for point cloud classification and segmentation, discussing their design, performance, challenges, and future directions in 3D vision tasks.

Contribution

It provides a comprehensive categorization, performance evaluation, and critical analysis of deep learning models for point cloud tasks, highlighting architectural innovations and limitations.

Findings

Performance benchmarks of various architectures are compared.

Architectural innovations improve accuracy and efficiency.

Open challenges and future research directions are identified.

Abstract

Point cloud stands as the most widely adopted format for representing 3D shapes and scenes due to its simplicity and geometric fidelity. However, its inherent unordered and irregular nature, exacerbated by sensor noise and occlusions, introduces unique challenges for machine learning based methodologies. To combat these issues, diverse strategies have been developed, including converting to a format that has orderliness, extracting local geometry, and permutation-invariant or self-attention-based processing. In this paper, our focus is directed towards deep learning models for three fundamental tasks in 3D vision: point cloud classification, part segmentation, and semantic segmentation. We begin by formally defining point cloud data, followed by an in-depth discussion on its structural characteristics. Then, we categorize notable works based on their backbone structure and evaluate their performance on popular benchmarks. Beyond empirical comparison, we offer insights into architectural innovations and limitations. We also outline open challenges and promising future directions for 3D point cloud understanding.