Point-Voxel Adaptive Feature Abstraction for Robust Point Cloud Classification
This addresses the problem of handling real-world corruptions in point cloud data for applications like autonomous driving or robotics, representing an incremental improvement over existing methods.
The paper tackles robust point cloud classification under corruptions like occlusion and noise by proposing a Point-Voxel Adaptive (PV-Ada) feature abstraction framework, achieving an Overall Accuracy of 0.865 on ModelNet-C and ranking 2nd in a 2022 challenge.
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.