# Unsupervised Point Cloud Registration with Self-Distillation

arXiv: https://arxiv.org/abs/2409.07558 · HTML: https://mlsanity.com/paper/2409.07558/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Detection and dense prediction distillation, one of 12 directions in Knowledge distillation.
- Builds on 19 of its 34 references that are in the map.
- First posted on arXiv 11 Sep 2024.

## Placed in

- Knowledge distillation / [Detection and dense prediction distillation](https://mlsanity.com/distill/detection-and-dense-prediction-distillation/)

## Builds on (most cited first, 19 of 19)

- [Emerging Properties in Self-Supervised Vision Transformers](https://mlsanity.com/paper/2104.14294/) (2021)
- [Bootstrap your own latent: A new approach to self-supervised Learning](https://mlsanity.com/paper/2006.07733/) (2020)
- [nuScenes: A multimodal dataset for autonomous driving](https://mlsanity.com/paper/1903.11027/) (2019)
- [Scalability in Perception for Autonomous Driving: Waymo Open Dataset](https://mlsanity.com/paper/1912.04838/) (2019)
- [Self-training with Noisy Student improves ImageNet classification](https://mlsanity.com/paper/1911.04252/) (2019)
- [3DMatch: Learning Local Geometric Descriptors from RGB-D Reconstructions](https://mlsanity.com/paper/1603.08182/) (2016)
- [DSAC - Differentiable RANSAC for Camera Localization](https://mlsanity.com/paper/1611.05705/) (2016)
- [Geometric Transformer for Fast and Robust Point Cloud Registration](https://mlsanity.com/paper/2202.06688/) (2022)
- [PPFNet: Global Context Aware Local Features for Robust 3D Point Matching](https://mlsanity.com/paper/1802.02669/) (2018)
- [Learning multiview 3D point cloud registration](https://mlsanity.com/paper/2001.05119/) (2020)
- [SC^2-PCR: A Second Order Spatial Compatibility for Efficient and Robust Point Cloud Registration](https://mlsanity.com/paper/2203.14453/) (2022)
- [Generalized Differentiable RANSAC](https://mlsanity.com/paper/2212.13185/) (2022)
- [Bootstrap Your Own Correspondences](https://mlsanity.com/paper/2106.00677/) (2021)
- [Unsupervised Deep Probabilistic Approach for Partial Point Cloud Registration](https://mlsanity.com/paper/2303.13290/) (2023)
- [Self-supervised Geometric Perception](https://mlsanity.com/paper/2103.03114/) (2021)
- [Reliable Inlier Evaluation for Unsupervised Point Cloud Registration](https://mlsanity.com/paper/2202.11292/) (2022)
- [Extend Your Own Correspondences: Unsupervised Distant Point Cloud Registration by Progressive Distance Extension](https://mlsanity.com/paper/2403.03532/) (2024)
- [Self-Supervised Point Cloud Registration with Deep Versatile Descriptors](https://mlsanity.com/paper/2201.10034/) (2022)
- [Cascading Feature Extraction for Fast Point Cloud Registration](https://mlsanity.com/paper/2110.12204/) (2021)

## Built on by (most cited first, 4 of 4)

- [PseudoMapTrainer: Learning Online Mapping without HD Maps](https://mlsanity.com/paper/2508.18788/) (2025)
- [Towards Foundation Models for 3D Scene Understanding: Instance-Aware Self-Supervised Learning for Point Clouds](https://mlsanity.com/paper/2603.25165/) (2026)
- [3D Registration in 30 Years: A Survey](https://mlsanity.com/paper/2412.13735/) (2024)
- [NeuralBoneReg: An Instance-Specific Label-Free Point Cloud-Based Method for Multi-Modal Bone Surface Registration](https://mlsanity.com/paper/2511.14286/) (2025)

## Abstract (arXiv)

Rigid point cloud registration is a fundamental problem and highly relevant in robotics and autonomous driving. Nowadays deep learning methods can be trained to match a pair of point clouds, given the transformation between them. However, this training is often not scalable due to the high cost of collecting ground truth poses. Therefore, we present a self-distillation approach to learn point cloud registration in an unsupervised fashion. Here, each sample is passed to a teacher network and an augmented view is passed to a student network. The teacher includes a trainable feature extractor and a learning-free robust solver such as RANSAC. The solver forces consistency among correspondences and optimizes for the unsupervised inlier ratio, eliminating the need for ground truth labels. Our approach simplifies the training procedure by removing the need for initial hand-crafted features or consecutive point cloud frames as seen in related methods. We show that our method not only surpasses them on the RGB-D benchmark 3DMatch but also generalizes well to automotive radar, where classical features adopted by others fail. The code is available at https://github.com/boschresearch/direg .