The task of point cloud classification suffers from the problem of insufficient data, and data augmentation is an effective method to alleviate this problem. However, the effect of conventional geometric-based point cloud augmentation strategies is insufficient. In mix-based augmentation strategies, the proportion of points is used as the weight for soft labels, which is not reasonable as the number of points does not always accurately represent the significance of features.

To solve the problems, a research team led by Xianghua YING published their new research on 12 Mar 2024 in Frontiers of Computer Science co-published by Higher Education Press and Springer Nature.

The team proposes a novel mix-based point cloud augmentation strategy called FPSMix, and a novel significance-based loss for point cloud classification. These methods achieve comparable classification accuracy to state-of-the-art methods, and can be used in conjunction with other data augmentation methods, such as conventional data augmentation and PointCutMix. Extensive ablation studies demonstrate that the proposed approach can be seamlessly combined with existing methods to enhance the robustness of the baseline model.

DOI: 10.1007/s11704-023-3455-4