A Collection of DL-based Point Cloud Registration Methods

This repository provides a summary of deep learning based point cloud registration algorithms.

If you find this repository helpful, we would greatly appreciate it if you could cite our paper: https://doi.org/10.24963/ijcai.2024/922 and http://arxiv.org/abs/2404.13830.

@inproceedings{ijcai2024p922,
  title     = {A Comprehensive Survey and Taxonomy on Point Cloud Registration Based on Deep Learning},
  author    = {Zhang, Yu-Xin and Gui, Jie and Cong, Xiaofeng and Gong, Xin and Tao, Wenbing},
  booktitle = {Proceedings of the Thirty-Third International Joint Conference on
               Artificial Intelligence, {IJCAI-24}},
  publisher = {International Joint Conferences on Artificial Intelligence Organization},
  pages     = {8344--8353},
  year      = {2024},
  month     = {8}
}

@misc{zhang2025deeplearningbasedpointcloud,
      title={Deep Learning-Based Point Cloud Registration: A Comprehensive Survey and Taxonomy}, 
      author={Yu-Xin Zhang and Jie Gui and Baosheng Yu and Xiaofeng Cong and Xin Gong and Wenbing Tao and Dacheng Tao},
      year={2025},
      eprint={2404.13830},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2404.13830}, 
}

We classify registration algorithms into supervised and unsupervised, as follows.

Supervised Point Cloud Registration Methods

• 1. Registration Procedure
  • 1. 1 Descriptor Extraction
  • 1. 2 Overlap Prediction
  • 1. 3 Similarity Matrix Optimization
  • 1. 4 Outlier Filtering
  • 1. 5 Transformation Parameter Estimation
  • 1. 6 Others
• 2. Optimization Strategy
  • 2. 1 GMM-Based
  • 2. 2 Bayesian-Based
  • 2. 3 Diffusion-Based
  • 2. 4 Multimodality-Based
  • 2. 5 Pretrain-Based
• 3. Learning Paradigm
  • 3. 1 Contrastive Learning
  • 3. 2 Meta Learning
  • 3. 3 Reinforcement Learning
• 4. Network Enhancement
  • 4. 1 Convolution-Based
  • 4. 2 Transformer-Based
• 5. Integration of Traditional Algorithms
  • 5. 1 Iterative Closest Point
  • 5. 2 Robust Point Matching
  • 5. 3 Lucas-Kanade

Unsupervised Point Cloud Registration Methods

• 1. Correspondence-free
  • 1. 1 One-stage Registration
  • 1. 2 Iterative Registration
• 2. Correspondence-based
  • 2. 1 RGB-D
  • 2. 2 Probability Model
  • 2. 3 Descriptor-Based
  • 2. 4 Geometric Consistency-Based

Datasets

Supervised Point Cloud Registration Methods

1. Registration Procedure

1. 1 Descriptor Extraction

1.1.1 Point-based

• D3Feat: Joint Learning of Dense Detection and Description of 3D Local Features [paper] [code] (2020, CVPR)

• GMCNet: Robust Partial-to-Partial Point Cloud Registration in a Full Range [paper] [code] (2024, RAL)

• Deep Semantic Graph Matching for Large-Scale Outdoor Point Cloud Registration [paper] [code] (2024, TGRS)

• RoCNet++: Triangle-based Descriptor for Accurate and Robust Point Cloud Registration [paper] [code] (2024, Pattern Recognition)

• PARE-Net: Position-Aware Rotation-Equivariant Networks for Robust Point Cloud Registration [paper] [code] (2025, ECCV)

1.1.2 Patch-based

• 3DMatch: Learning Local Geometric Descriptors from RGB-D Reconstructions [paper] [code] (2017, CVPR)

• PPFNet: Global Context Aware Local Features for Robust 3D Point Matching [paper] [code] (2018, CVPR)

• You Only Hypothesize Once: Point Cloud Registration with Rotation-equivariant Descriptors. [paper] [code] (2022, ACM MM)

• RoReg: Pairwise Point Cloud Registration with Oriented Descriptors and Local Rotations. [paper] [code] (2023, PAMI)

• Distinctive 3D local deep descriptors. [paper] [code] (2021, ICPR)

• GeDi: Learning General and Distinctive 3D Local Deep Descriptors for Point Cloud Registration. [paper] [code] (2022, PAMI)

• HA-TiNet: Learning a Distinctive and General 3D Local Descriptor for Point Cloud Registration [paper] [code] (2024, TVCG)

1.1.2 Voxel-based

• The Perfect Match: 3D Point Cloud Matching With Smoothed Densities. [paper] [code] (2019, CVPR)

• SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration. [paper] [code] (2021, CVPR)

• SphereNet: Learning a Noise-Robust and General Descriptor for Point Cloud Registration [paper] [code] (2023, TGRS)

1. 2 Overlap Prediction

• PREDATOR: Registration of 3D Point Clouds with Low Overlap. [paper] [code] (2021, CVPR)

• OMNet: Learning Overlapping Mask for Partial-to-Partial Point Cloud Registration. [paper] [code] (2021, ICCV)

• RORNet: Partial-to-Partial Registration Network With Reliable Overlapping Representations [paper] [code] (2024, TNNLS)

• STORM: Structure-Based Overlap Matching for Partial Point Cloud Registration. [paper] [code] (2023, TPAMI)

• A Unified BEV Model for Joint Learning of 3D Local Features and Overlap Estimation [paper] [code] (2023, ICRA)

• Low Overlapping Point Cloud Registration Using Mutual Prior Based Completion Network [paper] [code] (2024, TIP)

1. 3 Similarity Matrix Optimization

• PRNet: Self-supervised Learning for Partial-to-partial Registration. [paper] [code] (2019, NIPS)

• FIRE-Net: Feature Interactive Representation for Point Cloud Registration. [paper] [code] (2021, ICCV)

• One-Inlier is First: Towards Efficient Position Encoding for Point Cloud Registration. [paper] [code] (2022, NIPS)

• End-to-end Learning the Partial Permutation Matrix for Robust 3D Point Cloud Registration. [paper] [code] (2022, AAAI)

1. 4 Outlier Filtering

• Deep Hough Voting for Robust Global Registration. [paper] [code] (2021, ICCV)

• DLF: Partial Point Cloud Registration With Deep Local Feature. [paper] [code] (2023, TAI)

• PointDSC: Robust Point Cloud Registration using Deep Spatial Consistency. [paper] [code] (2021, CVPR)

• SC2-PCR: A Second Order Spatial Compatibility for Efficient and Robust Point Cloud Registration. [paper] [code] (2022, CVPR)

• SC-PCR++: Rethinking the Generation and Selection for Efficient and Robust Point Cloud Registration [paper] [code] (2023, TPAMI)

• 3D Registration with Maximal Cliques. [paper] [code] (2023, CVPR)

• MAC: Maximal Cliques for 3D Registration. [paper] [code] (2024, TPAMI)

• Hunter: Exploring High-Order Consistency for Point Cloud Registration With Severe Outliers. [paper] [code] (2023, TPAMI)

• Robust Point Cloud Registration via Random Network Co-Ensemble. [paper] [code] (2024, TCSVT)

• Scalable 3D Registration via Truncated Entry-wise Absolute Residuals. [paper] [code] (2024, CVPR)

• FastMAC: Stochastic Spectral Sampling of Correspondence Graph. [paper] [code] (2024, CVPR)

1. 5 Transformation Parameter Estimation

• Self-supervised Rigid Transformation Equivariance for Accurate 3D Point Cloud Registration [paper] [code] (2022, PR)

• DeTarNet: Decoupling Translation and Rotation by Siamese Network for Point Cloud Registration. [paper] [code] (2022, AAAI)

• FINet: Dual Branches Feature Interaction for Partial-to-Partial Point Cloud Registration [paper] [code] (2022, AAAI)

• Learning Compact Transformation Based on Dual Quaternion for Point Cloud Registration [paper] [code] (2024, TIM)

• Q-reg: End-to-end trainable point cloud registration with surface curvature [paper] [code] (2024, 3DV)

1. 6 Others

• DeepVCP: An End-to-End Deep Neural Network for Point Cloud Registration. [paper] [code] (2019, ICCV)

• HRegNet: A Hierarchical Network for Large-scale Outdoor LiDAR Point Cloud Registration. [paper] [code] (2021, ICCV)

• Robust Point Cloud Registration Framework Based on Deep Graph Matching [paper] [code] (2021, CVPR)

• BUFFER: Balancing Accuracy, Efficiency, and Generalizability in Point Cloud Registration [paper] [code] (2023, CVPR)

• Learning and Matching Multi-View Descriptors for Registration of Point Clouds. [paper] [code] (2018, ECCV)

• Learning Multiview 3D Point Cloud Registration [paper] [code] (2020, CVPR)

• Robust Multiview Point Cloud Registration with Reliable Pose Graph Initialization and History Reweighting. [paper] [code] (20223, CVPR)

2. Optimization Strategy

2.1 GMM-Based

• DeepGMR: Learning Latent Gaussian Mixture Models for Registration. [paper] [code] (2020, ECCV)

• OGMM: Overlap-guided Gaussian Mixture Models for Point Cloud Registration. [paper] [code] (2023, WACV)

• Point Cloud Registration Based on Learning Gaussian Mixture Models With Global-Weighted Local Representations. [paper] [code] (2023, GRSL)

2.2 Bayesian-Based

• VBReg: Robust Outlier Rejection for 3D Registration with Variational Bayes. [paper] [code] (2023, CVPR)

2.3 Diffusion-Based

• Point Cloud Registration With Zero Overlap Rate and Negative Overlap Rate [paper] [code] (2023, RAL)

• PosDiffNet: Positional Neural Diffusion for Point Cloud Registration in a Large Field of View with Perturbations [paper] [code] (2024, CVPR)

• PointDifformer: Robust Point Cloud Registration with Neural Diffusion and Transformer [paper] [code] (2024, TGRS)

• Diff-Reg: Diffusion-Based Correspondence Searching in Doubly Stochastic Matrix Space for Point Cloud Registration [paper] [code] (2025, ECCV)

• Se(3) diffusion model-based point cloud registration for robust 6d object pose estimation [paper] [code] (2024, NIPS)

2.4 Multimodality-Based

• PCR-CG: Point Cloud Registration via Deep Explicit Color and Geometry [paper] [code] (2022, ECCV)

• ImLoveNet: Misaligned Image-supported Registration Network for Low-overlap Point Cloud Pairs [paper] [code] (2022, SIGGRAPH)

• IMFNet: Interpretable Multimodal Fusion for Point Cloud Registration. [paper] [code] (2022, RAL)

• GMF: General Multimodal Fusion Framework for Correspondence Outlier Rejection. [paper] [code] (2022, RAL)

• IGReg: Image-geometry-assisted Point Cloud Registration via Selective Correlation Fusion. [paper] [code] (2024, TMM)

• SemReg: Semantics Constrained Point Cloud Registration [paper] [code] (2025, ECCV)

2.5 Pretrain-Based.

• SIRA-PCR: Sim-to-Real Adaptation for 3D Point Cloud Registration. [paper] [code] (2023, ICCV)

• Zero-Shot Point Cloud Registration [paper] (2023)

• PointRegGPT: Boosting 3D Point Cloud Registration using Generative Point-Cloud Pairs for Training [paper] [code] (2025, ECCV)

• Boosting 3D Point Cloud Registration by Transferring Multi-modality Knowledge [paper] [code] (2023, ICRA)

3. Learning Paradigm

3.1 Contrastive Learning

• PointCLM: A Contrastive Learning-based Framework for Multi-instance Point Cloud Registration [paper] [code] (2022, ECCV)

• SCRnet: A Spatial Consistency Guided Network using Contrastive Learning for Point Cloud Registration [paper] [code]

• Density-invariant Features for Distant Point Cloud Registration [paper] [code] (2023, ICCV)

UMERegRobust: Universal Manifold Embedding Compatible Features for Robust Point Cloud Registration [paper] [code] (2025, ECCV)

3.2 Meta Learning

• Point-TTA: Test-Time Adaptation for Point Cloud Registration Using Multitask Meta-Auxiliary Learning [paper] (2023, ICCV)

• 3D Meta-Registration: Learning to Learn Registration of 3D Point Clouds [paper] (2020)

3.3 Reinforcement Learning

• Reagent: Point cloud registration using imitation and reinforcement learning [paper] (2021, CVPR)

• Point Cloud Registration via Heuristic Reward Reinforcement Learning [paper] (2023)

4. Network Enhancement

4.1 Convolution-Based

• Fully convolutional geometric features [paper] [code] (2019, ICCV)

• Deep Global Registration [paper] [code] (2020, CVPR)

• 3DRegNet: A Deep Neural Network for 3D Point Registration. [paper] [code] (2020, CVPR)

• Accelerating Point Cloud Registration with Low Overlap using Graphs and Sparse Convolutions [paper] [code] (2023, TMM)

• PCAM: Product of Cross-Attention Matrices for Rigid Registration of Point Clouds. [paper] [code] (2021, ICCV)

• SACF-Net: Skip-Attention Based Correspondence Filtering Network for Point Cloud Registration [paper] (2023, TCSVT)

4.2 Transformer-based

• GeoTransformer: Fast and Robust Point Cloud Registration With Geometric Transformer. [paper] [code] (2023, PAMI)

• SPEAL: Skeletal Prior Embedded Attention Learning for Cross-Source Point Cloud Registration [paper] (2024, AAAI)

• RoITr: Rotation-Invariant Transformer for Point Cloud Matching. [paper] [code] (2023, CVPR)

• PEAL: Prior-embedded Explicit Attention Learning for Low-overlap Point Cloud Registration. [paper] [code] (2023, CVPR)

• EGST: Enhanced Geometric Structure Transformer for Point Cloud Registration [paper] (2024, TVCG)

• Full Transformer Framework for Robust Point Cloud Registration With Deep Information Interaction [paper] [code] (2023, TNNLS)

• Neighborhood Multi-Compound Transformer for Point Cloud Registration [paper] (2024, TCSVT)

• A Consistency-Aware Spot-Guided Transformer for Versatile and Hierarchical Point Cloud Registration [paper] [code] (2024, NIPS)

Learning Instance-Aware Correspondences for Robust Multi-Instance Point Cloud Registration in Cluttered Scenes [paper] [code] (2024, CVPR)

• Dynamic Cues-assisted Transformer for Robust Point Cloud Registration [paper] [code] (2024, CVPR)

• REGTR: End-to-end Point Cloud Correspondences with Transformers. [paper] [code] (2022, CVPR)

• RegFormer: An Efficient Projection-Aware Transformer Network for Large-Scale Point Cloud Registration [paper] [code] (2023, ICCV)

5. Integration of Traditional Algorithms

5.1 Iterative Closest Point

• Dcpcr: Deep compressed point cloud registration in large-scale outdoor environments [paper] (2022, RAL)

• Deep Closest Point: Learning Representations for Point Cloud Registration. [paper] [code] (2019, ICCV)

• Global-PBNet: A Novel Point Cloud Registration for Autonomous Driving [paper] (2022, TITS)

• IDAM: Iterative Distance-Aware Similarity Matrix Convolution with Mutual-Supervised Point Elimination for Efficient Point Cloud Registration. [paper] [code] (2022, ECCV)

5.2 Robust Point Matching

• RPM-Net: Robust Point Matching using Learned Features [paper] [code] (2020, CVPR)

5.3 Lucas-Kanade

• PointNetLK: Robust & Efficient Point Cloud Registration Using PointNet. [paper] [code] (2019, CVPR)

• PointNetLK Revisited. [paper] [code] (2021, CVPR)

Unsupervised Point Cloud Registration Methods

1. Correspondence-free

1.1 One-stage Registration

• PPF-FoldNet: Unsupervised Learning of Rotation Invariant 3D Local Descriptors. [paper] [code] (2018, ECCV)

• 3D point cloud registration with multi-scale architecture and unsupervised transfer learning. [paper] [code] (2021, 3DV)

• UPCR: A Representation Separation Perspective to Correspondence-Free Unsupervised 3-D Point Cloud Registration. [paper] (2023, GRSL)

• UGMM: Unsupervised Point Cloud Registration by Learning Unified Gaussian Mixture Models. [paper] [code] (2022, RAL)

1.2 Iterative Registration

• Deep-3DAligner: Unsupervised 3D Point Set Registration Network With Optimizable Latent Vector [paper] (2020)

• Research and Application on Cross-source Point Cloud Registration Method Based on Unsupervised Learning。 [paper] (2023, CYBER))

• Learning Discriminative Features via Multi-Hierarchical Mutual Information for Unsupervised Point Cloud Registration [paper] (2024, TCSVT)

• PCRNet: Point Cloud Registration Network using PointNet Encoding [paper] [code] (2019)

2. Correspondence-based

2.1 RGB-D

• UnsupervisedR&R: Unsupervised Point Cloud Registration via Differentiable Rendering [paper] [code] (2021, CVPR)

• PointMBF: A Multi-scale Bidirectional Fusion Network for Unsupervised RGB-D Point Cloud Registration [paper] [code] (2023, ICCV)

• Improving RGB-D Point Cloud Registration by Learning Multi-scale Local Linear Transformation [paper] [code] (2022, ECCV)

• NeRF-Guided Unsupervised Learning of RGB-D Registration [paper] [code] (2024)

• Discriminative correspondence estimation for unsupervised rgb-d point cloud registration [paper] [code] (2024, TCSVT)

2.2 Probability Model

• Overlap Bias Matching is Necessary for Point Cloud Registration [paper] (2023)

• UGMM: Unsupervised Point Cloud Registration by Learning Unified Gaussian Mixture Models. [paper] [code] (2022, RAL)

• UDPReg: Unsupervised Deep Probabilistic Approach for Partial Point Cloud Registration. [paper] [code] (2023, CVPR)

• Planning with Learned Dynamic Model for Unsupervised Point Cloud Registration [paper] (2021, IJCAI)

• CEMNet: Sampling Network Guided Cross-Entropy Method for Unsupervised Point Cloud Registration. [paper] [code] (2021, ICCV)

2.3 Descriptor-based

• DeepUME: Learning the Universal Manifold Embedding for Robust Point Cloud Registration [paper] [code] (2021, BMVC)

• GTINet: Global Topology-Aware Interactions for Unsupervised Point Cloud Registration [paper] [code] (2024, TCSVT)

• Corrnet3d: Unsupervised end-to-end learning of dense correspondence for 3d point clouds [paper] [code] (2021, CVPR)

• R-PointHop: A Green, Accurate, and Unsupervised Point Cloud Registration Method [paper] [code] (2022, TIP)

2.4 Geometric Consistency-based

• RIENet: Reliable Inlier Evaluation for Unsupervised Point Cloud Registration. [paper] [code] (2022, AAAI)

• RegiFormer: Unsupervised Point Cloud Registration via Geometric Local-to-Global Transformer and Self Augmentation [paper] (2024, TGRS)

• Extend Your Own Correspondences: Unsupervised Distant Point Cloud Registration by Progressive Distance Extension [paper] [code] (2024, CVPR)

• Inlier Confidence Calibration for Point Cloud Registration [paper] (2024, CVPR)

• Mining and Transferring Feature-Geometry Coherence for Unsupervised Point Cloud Registration [paper] [code] (2024, NIPS)

Datasets

• ETH: Challenging Data Sets for Point Cloud Registration Algorithms. [paper] [code]

• KITTI: Are We Ready for Autonomous Driving? The KITTI Vision Benchmark Suite. [paper] [code]

• ICL-NUIM: A Benchmark for RGB-D Visual Odometry, 3D Reconstruction and SLAM. [paper] [code]

• ModelNet40: 3D ShapeNets: A Deep Representation for Volumetric Shapes. [paper] [code]

• ShapeNet: An Information-Rich 3D Model Repository. [paper] [code]

• RedWood: A large dataset of object scans [paper] [code]

• 3DMatch: Learning the Matching of Local 3D Geometry in Range Scans. [paper] [code]

• Oxford RobotCar: 1 year, 1000 km: The oxford robotcar dataset [paper] [code]

• ScanObjectNN. Revisiting Point Cloud Classification: A New Benchmark Dataset and Classification Model on Real-World Data. [paper] [code]

• WHU-TLS: Registration of Large-scale Terrestrial Laser Scanner Point Clouds: A Review and Benchmark. [paper] [code]

• Nuscenes: A multimodal dataset for autonomous driving [paper] [code]

• MVP-RG: Robust partial-to-partial point cloud registration in a full rang [paper] [code]

• FlyingShapes: SIRA-PCR: Sim-to-Real Adaptation for 3D Point Cloud Registration. [paper] [code]