[ICRA 2024] Benchmarking Classical and Learning-Based Multibeam Point Cloud Registration*

The data is now released and can be found here. This repository accompanies the ICRA 2024 paper:

Benchmarking Classical and Learning-Based Multibeam Point Cloud Registration

Li Ling¹, Jun Zhang², Nils Bore³, John Folkesson¹, Anna Wåhlin⁴

|¹KTH Royal Institute of Technology|²TU Graz|³Ocean Infinity|⁴University of Gothenburg|

For more information, please check out the project website

Contacts

If you have any questions, feel free to contact us at:

• Li Ling (liling@kth.se)

Instructions

This repository contains the implementation for the MBES Dataset class and data loader, the classical methods GICP and FPFH, as well as the code for metrics computation and evaluations.

The code use to run the learning-based models are found in the following repository forks:

• FCGF
• DGR
• Predator
• BathyNN

The dataset, pretrained models and evaluation results can be found here. Note that the dataset only contains the patches segmented according to the paper description. To construct a registration dataset, please consult main.py. If you want the exact data pairs and transforms as used in the paper, you can also extract these from the npz files containing in each method's evaluation results.

Citation

If you find this code useful for your work, please consider citing:

@inproceedings{ling2024benchmarking,
            title={Benchmarking Classical and Learning-Based Multibeam Point Cloud Registration}, 
            author={Ling, Li and Zhang, Jun and Bore, Nils and Folkesson, John and Wåhlin, Anna},
            booktitle={2024 IEEE International Conference on Robotics and Automation (ICRA)},
            year={2024},
            organization={IEEE}
}

Acknowledgements

In this project, we use part of the official implementations of the following work:

• FCGF
• DGR
• Predator
• BathyNN

We thank the respective authors for open sourcing their work.