End-to-End Learning Local Multi-view Descriptors for 3D Point Clouds
About
In this work, we propose an end-to-end framework to learn local multi-view descriptors for 3D point clouds. To adopt a similar multi-view representation, existing studies use hand-crafted viewpoints for rendering in a preprocessing stage, which is detached from the subsequent descriptor learning stage. In our framework, we integrate the multi-view rendering into neural networks by using a differentiable renderer, which allows the viewpoints to be optimizable parameters for capturing more informative local context of interest points. To obtain discriminative descriptors, we also design a soft-view pooling module to attentively fuse convolutional features across views. Extensive experiments on existing 3D registration benchmarks show that our method outperforms existing local descriptors both quantitatively and qualitatively.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Point cloud registration | 3DMatch (test) | -- | 339 | |
| Feature Matching | 3DMatch (Origin) | STD2.8 | 33 | |
| Feature Matching | ETH dataset (test) | FMR (Gazebo Summer)85.3 | 23 | |
| Descriptor matching | 3DMatch Rotated | -- | 18 | |
| 3D local descriptor matching | 3DMatch | Average Recall97.5 | 16 | |
| Feature Matching | 3DMatch | FMR (tau_2=0.05)97.5 | 15 | |
| Point cloud registration | 3DMatch Origin v1 | Registration Recall82.5 | 12 | |
| Feature Matching | 3DMatch Rotated (test) | FMR0.969 | 12 | |
| Point cloud registration | 3DLoMatch v1 (Origin) | Registration Recall41.3 | 11 | |
| Point cloud registration | 3DLoMatch Rotated v1 | Registration Recall40.1 | 11 |