[Paper Review] Advancements in Point Cloud Data Augmentation for Deep Learning: A Survey

Qinfeng Zhu, Lei Fan+1 more|arXiv (Cornell University)|Aug 23, 2023

3D Shape Modeling and Analysis154 references4 citations

TL;DR

This survey presents a comprehensive taxonomy of point cloud data augmentation methods for deep learning, categorizing them into basic and specialized techniques. It evaluates their effectiveness in classification, segmentation, and detection tasks, offering guidance on method selection and identifying future research directions for improved generalization and robustness in low-data regimes.

ABSTRACT

Deep learning (DL) has become one of the mainstream and effective methods for point cloud analysis tasks such as detection, segmentation and classification. To reduce overfitting during training DL models and improve model performance especially when the amount and/or diversity of training data are limited, augmentation is often crucial. Although various point cloud data augmentation methods have been widely used in different point cloud processing tasks, there are currently no published systematic surveys or reviews of these methods. Therefore, this article surveys these methods, categorizing them into a taxonomy framework that comprises basic and specialized point cloud data augmentation methods. Through a comprehensive evaluation of these augmentation methods, this article identifies their potentials and limitations, serving as a useful reference for choosing appropriate augmentation methods. In addition, potential directions for future research are recommended. This survey contributes to providing a holistic overview of the current state of point cloud data augmentation, promoting its wider application and development.

Motivation & Objective

• To address the lack of systematic reviews on point cloud data augmentation methods in deep learning.
• To propose a structured taxonomy that categorizes augmentation techniques into basic and specialized methods based on complexity and application context.
• To evaluate the performance and limitations of existing augmentation methods across key point cloud tasks such as classification, segmentation, and detection.
• To provide researchers with actionable guidance on selecting appropriate augmentation strategies for specific tasks and data constraints.
• To identify open challenges and recommend future research directions in point cloud data augmentation.

Proposed method

• The authors propose a two-tier taxonomy: basic augmentation (e.g., geometric transformations like rotation, scaling, flipping) and specialized augmentation (e.g., mixup, cutout, generative methods) tailored for specific challenges.
• The survey systematically reviews 68 recent studies (2017–2023) to identify and classify augmentation techniques used in point cloud processing.
• Basic augmentation methods are evaluated for their simplicity, versatility, and widespread use in combination with one another.
• Specialized methods are analyzed based on their design goals—such as improving generalization, handling class imbalance, or enabling self-supervised learning.
• The framework integrates insights from image data augmentation (e.g., Mixup, CutMix) and adapts them to 3D point cloud characteristics.
• The authors use qualitative and comparative analysis to assess method performance, computational cost, and applicability across different point cloud tasks.

Experimental results

Research questions

• RQ1 What are the key categories and subcategories of point cloud data augmentation methods in deep learning?
• RQ2 How do basic and specialized augmentation techniques differ in terms of design, complexity, and application scope?
• RQ3 What are the performance trade-offs and limitations of existing augmentation methods in point cloud classification, segmentation, and detection?
• RQ4 How do augmentation techniques contribute to model generalization and robustness under low-data or imbalanced data conditions?
• RQ5 What are the promising future research directions for advancing point cloud data augmentation?

Key findings

• The survey identifies that basic augmentation methods—such as rotation, scaling, and flipping—are widely used due to their simplicity and effectiveness in increasing data diversity.
• Specialized methods like PointMixUp, CutMix, and SAGE-Mix show improved generalization and robustness in point cloud classification by learning from mixed or masked samples.
• Generative augmentation techniques such as GANs and diffusion models (e.g., PointGAN, Diffusion models) demonstrate potential for synthesizing realistic point clouds but face challenges in training stability and mode collapse.
• Auto-augmentation frameworks like PointAugment and MaxUp show promise in automatically discovering optimal augmentation policies, reducing hyperparameter sensitivity.
• Despite progress, many methods remain sensitive to data distribution shifts and lack consistent performance across diverse point cloud benchmarks.
• The study highlights that current augmentation methods often overlook structural and semantic consistency, suggesting a need for more semantically aware augmentation strategies.