Research map / Point clouds / Registration and matching
Deep Closest Point: Learning Representations for Point Cloud Registration
Yue Wang, Justin M. Solomon
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use DGCNN instead of PointNet embedding (ablation) | DCP-v2 · ModelNet40 (unseen point clouds) | RMSE(R) degrees: 7.061375 -> 1.143385 | Improved |
| use DGCNN instead of PointNet embedding (ablation) | DCP-v1 · ModelNet40 (unseen point clouds) | RMSE(R) degrees: 4.124127 -> 2.545697 | Improved |
| use SVD layer instead of MLP to output rigid transform (ablation) | DCP-v2 · ModelNet40 (unseen point clouds) | RMSE(R) degrees: 3.150191 -> 1.143385 | Improved |
| use SVD layer instead of MLP to output rigid transform (ablation) | DCP-v1 · ModelNet40 (unseen point clouds) | RMSE(R) degrees: 4.595206 -> 2.545697 | Improved |
| add Transformer attention module (DCP-v1 to DCP-v2) (ablation) | DCP · ModelNet40 (unseen point clouds) | RMSE(R) degrees: 2.545697 -> 1.143385 | Improved |
| add Transformer attention module (DCP-v1 to DCP-v2) (ablation) | DCP · ModelNet40 (unseen categories) | RMSE(R) degrees: 4.381938 -> 3.150191 | Improved |
| increase embedding dimension from 512 to 1024 (ablation) | DCP-v2 · ModelNet40 (unseen point clouds) | RMSE(R) degrees: 1.143385 -> 1.103424 | Improved |
| increase embedding dimension from 512 to 1024 (ablation) | DCP-v1 · ModelNet40 (unseen point clouds) | RMSE(R) degrees: 2.545697 -> 2.700225 | Hurt |
| initialize ICP with DCP output | ICP · ModelNet40 (unseen point clouds) | RMSE(R) degrees | Improved |
| use DCP-v2 instead of ICP | rigid registration · ModelNet40 (unseen point clouds) | RMSE(R) degrees: 29.914835 -> 1.143385 | Improved |
| evaluate on unseen categories instead of unseen point clouds | DCP-v2 · ModelNet40 | RMSE(R) degrees: 1.143385 -> 3.150191 | Hurt |
| add Gaussian noise to test point clouds | FGR · ModelNet40 | RMSE(R) degrees: 9.362772 -> 24.651468 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PointNetLK — PointNetLK: Robust & Efficient Point Cloud Registration Using PointNet (CVPR 2019) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- RPM-Net: Robust Point Matching Using Learned Features (CVPR 2020 · similarity 0.93)
- DeepPRO: Deep Partial Point Cloud Registration of Objects (ICCV 2021 · similarity 0.91)
- Robust Point Cloud Registration Framework Based on Deep Graph Matching (CVPR 2021 · similarity 0.91)
- PointMC: Multi-instance Point Cloud Registration based on Maximal Cliques (ICML 2024 · similarity 0.89)
- FreeReg: Image-to-Point Cloud Registration Leveraging Pretrained Diffusion Models and Monocular Depth Estimators (ICLR 2024 · similarity 0.89)
- PosDiffNet: Positional Neural Diffusion for Point Cloud Registration in a Large Field of View with Perturbations (AAAI 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
