PRNet: Self-Supervised Learning for Partial-to-Partial Registration
About
We present a simple, flexible, and general framework titled Partial Registration Network (PRNet), for partial-to-partial point cloud registration. Inspired by recently-proposed learning-based methods for registration, we use deep networks to tackle non-convexity of the alignment and partial correspondence problems. While previous learning-based methods assume the entire shape is visible, PRNet is suitable for partial-to-partial registration, outperforming PointNetLK, DCP, and non-learning methods on synthetic data. PRNet is self-supervised, jointly learning an appropriate geometric representation, a keypoint detector that finds points in common between partial views, and keypoint-to-keypoint correspondences. We show PRNet predicts keypoints and correspondences consistently across views and objects. Furthermore, the learned representation is transferable to classification.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Point cloud registration | ModelNet40 (Unseen categories) | RMSE (Rotation)3.199 | 36 | |
| Point cloud registration | ModelNet40 RPMNet manner (Unseen Shapes) | RMSE(R)1.588 | 32 | |
| Point cloud registration | ModelNet40 twice-sampled (TS) unseen categories (test) | RMSE (Rotation)2.712 | 30 | |
| Point cloud registration | ModelNet40 Unseen Categories with Gaussian Noise RPMNet manner (OS) | RMSE (Rotation)1.911 | 21 | |
| Point cloud registration | ModelNet40 PRNet generation manner with Gaussian noise (unseen categories) | RMSE (Rotation)3.241 | 20 | |
| Point cloud registration | ModelNet40 Gaussian Noise twice-sampled (test) | RMSE (R)3.197 | 11 | |
| Rigid Point Cloud Registration | ModelNet40 unseen objects + Gaussian noise | RMSE (R)4.32 | 7 | |
| Point cloud registration | ModelNet40 8 axisymmetrical categories PRNet/RPMNet generation (once-sampled (OS)) | RMSE (Rotation)5.979 | 4 | |
| Point cloud registration | ModelNet40 8 axisymmetrical categories PRNet/RPMNet generation (twice-sampled (TS)) | RMSE (R)13.773 | 4 |