PyTorch implementation of the paper: HECPG: Hyperbolic Embedding and Confident Patch-Guided Network for Point Cloud Matching.
The codes of our project mainly base on GeoTransformer, please follow it.
To train a model:
python trainval.py
To test a model:
python test.py --snapshot=weights_path --benchmark=3DMatch
python eval.py --benchmark=3DMatch --method=ransac
If you find our work useful in your research, please consider citing:
@article{xie2024hecpg,
title={HECPG: Hyperbolic Embedding and Confident Patch-Guided Network for Point Cloud Matching},
author={Xie, Yifan and Zhu, Jihua and Li, Shiqi and Hu, Naiwen and Shi, Pengcheng},
journal={IEEE Transactions on Geoscience and Remote Sensing},
year={2024},
volume={62},
number={},
pages={1-12},
publisher={IEEE}
}
This code is developed heavily relying on GeoTransformer, PREDATOR and hyp-ViT. Thanks for these great projects.