Yongheng Zhao, Tolga Birdal, Haowen Deng, Federico Tombari
In this paper, we propose 3D point-capsule networks, an auto-encoder designed to process sparse 3D point clouds while preserving spatial arrangements of the input data. 3D capsule networks arise as a direct consequence of our novel unified 3D auto-encoder formulation. Their dynamic routing scheme and the peculiar 2D latent space deployed by our approach bring in improvements for several common point cloud-related tasks, such as object classification, object reconstruction and part segmentation as substantiated by our extensive evaluations. Moreover, it enables new applications such as part interpolation and replacement.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| 3D | ModelNet40 | Classification Accuracy | 89.3 | 3D-PointCapsNet |
| Shape Representation Of 3D Point Clouds | ModelNet40 | Classification Accuracy | 89.3 | 3D-PointCapsNet |
| 3D Object Classification | ModelNet40 | Classification Accuracy | 89.3 | 3D-PointCapsNet |
| 3D Point Cloud Classification | ModelNet40 | Classification Accuracy | 89.3 | 3D-PointCapsNet |
| 3D Classification | ModelNet40 | Classification Accuracy | 89.3 | 3D-PointCapsNet |
| 3D Point Cloud Reconstruction | ModelNet40 | Classification Accuracy | 89.3 | 3D-PointCapsNet |