This repository is a reference implementation for "NM-Net: Mining Reliable Neighbors for Robust Feature Correspondences", CVPR 2019 oral. If you use this code in your research, please cite the paper.
@inproceedings{zhao2019nm,
title={NM-Net: Mining reliable neighbors for robust feature correspondences},
author={Zhao, Chen and Cao, Zhiguo and Li, Chi and Li, Xin and Yang, Jiaqi},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
pages={215--224},
year={2019}
}
pip install -r requirements.txt
Edit config.data_tr in config.py to prepare data for different datasets.
python ./dump_data.py
For the first time running main.py in each dataset, set the parameter initialize in Data_Loader to be True.
python ./main.py COLMAP #NARROW WIDE
An end-to-end version has been released which is independent to affine information. Please refer to code/NM_Net_v2.py.
The data processing and evaluation codes are borrowed from "Learning to Find Good Correspondences" (CVPR 2018). Please cite this paper if the code is useful for your research.
@inproceedings{moo2018learning,
title={Learning to find good correspondences},
author={Moo Yi, Kwang and Trulls, Eduard and Ono, Yuki and Lepetit, Vincent and Salzmann, Mathieu and Fua, Pascal},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
pages={2666--2674},
year={2018}
}