Point-Voxel Adaptive Feature Abstraction for Robust Point Cloud Classification
Lifa Zhu 11 Changwei Lin 11 Chen Zheng 22 Ninghua Yang 11
Abstract
Great progress has been made in point cloud classification with learning-based methods. However, complex scene and sensor inaccuracy in real-world application make point cloud data suffer from corruptions, such as occlusion, noise and outliers. In this work, we propose Point-Voxel based Adaptive (PV-Ada) feature abstraction for robust point cloud classification under various corruptions. Specifically, the proposed framework iteratively voxelize the point cloud and extract point-voxel feature with shared local encoding and Transformer. Then, adaptive max-pooling is proposed to robustly aggregate the point cloud feature for classification. Experiments on ModelNet-C dataset demonstrate that PV-Ada outperforms the state-of-the-art methods. In particular, we rank the $2^{nd}$ place in ModelNet-C classification track of PointCloud-C Challenge 2022, with Overall Accuracy (OA) being 0.865. Code will be available at https://github.com/zhulf0804/PV-Ada.
中文速览
真实场景中的点云数据常因遮挡、噪声和离群点等干扰而质量下降，导致现有分类模型鲁棒性不足。针对这一问题，研究者提出了PV-Ada（Point-Voxel based Adaptive feature abstraction）框架：先对点云做多尺度体素化（voxelization），再用共享权重的局部编码器与Transformer同时提取点级和体素级特征，捕捉从精细细节到粗糙轮廓的多层次信息；同时引入自适应最大池化（adaptive max-pooling），让模型学会为不同点分配重要性权重，从而压制噪声点和离群点的干扰。在专门评测点云鲁棒性的ModelNet-C数据集上，PV-Ada超越了所有已发表的最优方法，并在PointCloud-C Challenge 2022的分类赛道中以0.865的整体准确率获得第二名。这项工作表明，结合多尺度点-体素特征与自适应聚合策略，能够显著提升三维点云分类在复杂真实环境下的可靠性。
原文 arXiv:2210.15514;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2210.15514v2