(ICCV 2021) PyTorch implementation of Paper "Progressive Correspondence Pruning by Consensus Learning"
- [project page]
- [paper]
If you find the CLNet code useful, please consider citing:
@inproceedings{zhao2021progressive,
title={Progressive Correspondence Pruning by Consensus Learning},
author={Zhao, Chen and Ge, Yixiao and Zhu, Feng and Zhao, Rui and Li, Hongsheng and Salzmann, Mathieu},
booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision.},
year={2021}
}
Please start by installing the required libraries:
pip install -r requirements.txt
The code of this part is partially borrowed from [OANet] [CNe]. Please follow their instructions to download the training and testing data.
bash download_data.sh raw_data raw_data_yfcc.tar.gz 0 8 ## YFCC100M
tar -xvf raw_data_yfcc.tar.gz
bash download_data.sh raw_sun3d_test raw_sun3d_test.tar.gz 0 2 ## SUN3D
tar -xvf raw_sun3d_test.tar.gz
bash download_data.sh raw_sun3d_train raw_sun3d_train.tar.gz 0 63
tar -xvf raw_sun3d_train.tar.gz
After downloading the datasets, the initial matches can be generated by:
cd dump_match
bash yfcc.sh
bash sun3d.sh
The initial matches are generated over SIFT by default. The ones based on ORB and SuperPoint are also available by changing the settings of --suffix
and --desc_name
.
We provide a pretrained model on YFCC100M. The results in our paper can be reproduced by running the test script:
python ./test.py --use_ransac True --data_te ./data_dump/yfcc-sift-2000-test.hdf5 --output_dir ./logs/CLNet_yfcc_sift --model_path ./pretrained_models/clnet_yfcc_sift.pth
Please run the training script to train our model on YFCC100M after the data processing is done.
python ./train.py --data_tr ./data_dump/yfcc-sift-2000-train.hdf5
--data_te ./data_dump/yfcc-sift-2000-test.hdf5