PRNet: Self-Supervised Learning for Partial-to-Partial Registration

By Yue Wang, Justin M. Solomon. Neural Information Processing Systems, 2019

466 citations recorded by the source corpus.

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 problem. 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.

Subfields: Aerospace Engineering, Computational Mechanics, Geology

View Paper: https://www.semanticscholar.org/paper/a6cc088c2196040c08e62531d62cd7e7fae2ab28

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

1. Christopher Bongsoo Choy
2. Xiaoshui Huang
3. Guofeng Mei
4. Vladlen Koltun
5. Luca Carlone
6. Rana Abbas
7. Jian Zhang
8. Jingnan Shi

In Computational Mechanics

1. Christopher Bongsoo Choy
2. Xiaoshui Huang
3. Guofeng Mei
4. Rana Abbas
5. Jian Zhang
6. Vladlen Koltun
7. Zixin Luo
8. Xuyang Bai

In Geology

1. Christopher Bongsoo Choy
2. Vladlen Koltun
3. Xiaoshui Huang
4. Guofeng Mei
5. Wei Dong
6. Luca Carlone
7. Zixin Luo
8. Lei Zhou

Citations in the corpus

• A comprehensive survey on point cloud registration (2021, background): “Recent technologies [100], [36] propose keypoint-based solutions to solve partial overlap.”
• TEASER: Fast and Certifiable Point Cloud Registration (2020, background): “PRNet [83] extends DCP to aligning partially overlapping point clouds.”
• PREDATOR: Registration of 3D Point Clouds with Low Overlap (2020, methodology): “We failed to train PRNet [41] due to random crashes of the original code (also observed in [6]).”
• PointDSC: Robust Point Cloud Registration using Deep Spatial Consistency (2021, background): “Besides, end-toend registration networks [3, 62, 63, 71] have been proposed.”
• Geometric Transformer for Fast and Robust Point Cloud Registration (2022, unknown): “The first class [12,30,31,35] follows the idea of ICP [5], which iteratively establishes soft correspondences and computes the transformation with differentiable weighted SVD.”
• CoFiNet: Reliable Coarse-to-fine Correspondences for Robust Point Cloud Registration (2021, background): “The first [13, 14, 15] follows the idea of ICP [16], where they iteratively find dense correspondences and compute pose estimation.”
• TAPIR: Tracking Any Point with per-frame Initialization and temporal Refinement (2023, methodology): “The four ResNet layers consist of feature dimensions of [64, 128, 256, 256], with 2 ResNet blocks per layer.”
• RPM-Net: Robust Point Matching Using Learned Features (2020, unknown): “A concurrent work, PRNet [36] incorporates keypoint detection to handle partial visibility.”
• Deep Global Registration (2020, methodology): “We also tried to train PRNet [42] on our setup, but failed to get reasonable results due to random crashes and high-variance training losses.”
• VIME: Extending the Success of Self- and Semi-supervised Learning to Tabular Domain (2020, background): “Most existing works with pretext tasks are appropriate only for images or natural language: (i) surrogate classes prediction (scaling and translation) [11], (ii) rotation degree predictions [7], (iii) colorization [9], (iv) relative position of patches estimation [12], (v) jigsaw puzzle solving [8],”