Go to ICML 2025 Conference homepageRethinking Point Cloud Data Augmentation: Topologically Consistent Deformation
Data augmentation has been widely used in machine learning. Its main goal is to transform and expand the original data using various techniques, creating a more diverse and enriched training dataset. However, due to the disorder and irregularity of point clouds, existing methods struggle to enrich geometric diversity and maintain topological consistency, leading to imprecise point cloud understanding. In this paper, we propose SinPoint, a novel method designed to preserve the topological structure of the original point cloud through a homeomorphism. It utilizes the Sine function to generate smooth displacements. This simulates object deformations, thereby producing a rich diversity of samples. In addition, we propose a Markov chain Augmentation Process to further expand the data distribution by combining different basic transformations through a random process. Our extensive experiments demonstrate that our method consistently outperforms existing Mixup and Deformation methods on various benchmark point cloud datasets, improving performance for shape classification and part segmentation tasks. Specifically, when used with PointNet++ and DGCNN, our method achieves a state-of-the-art accuracy of 90.2 in shape classification with the real-world ScanObjectNN dataset. We release the code at https://github.com/CSBJian/SinPoint.
Existing data augmentation methods for point clouds struggled with maintaining both geometric diversity and topological consistency. This led to difficulties in understanding point clouds accurately, which hindered performance in tasks like shape classification and part segmentation.
We proposed a novel method called SinPoint, which uses a homeomorphism to preserve the topological structure of point clouds. The sine function was utilized to simulate smooth deformations, generating a diverse set of samples. Additionally, we introduced a Markov chain Augmentation Process to randomly combine various transformations and further enrich the data.
Our research improves the understanding of point clouds by creating more diverse and topologically consistent data, leading to better performance in shape classification and segmentation tasks. Meanwhile, our method provides a new perspective for point cloud data augmentation, constructing homeomorphic variants of the data to expand the distribution of the data.