Casacde-Stereo(CVPR2020 Oral)

Official source code of paper Cascade Cost Volume for High-Resolution Multi-View Stereo and Stereo Matching, arxiv

Installation

Requirements

• python 3.6
• Pytorch == 1.2 （The default interpolation of the high version is different from the low version, which will cause the results to be different）
• CUDA >= 9.0

pip install -r requirements.txt  --user

Option: Apex

install apex to enable synchronized batch normalization

git clone https://github.com/NVIDIA/apex.git
cd apex
pip install -v --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./

CasMVSNet(Multi View Stereo)

Refer to MVS-README.md

CasStereoNet(Stereo Matching)

Refer to Stereo-Matching-README.md

Citation

If you find this code useful in your research, please cite:

@inproceedings{gu2020cascade,
  title={Cascade cost volume for high-resolution multi-view stereo and stereo matching},
  author={Gu, Xiaodong and Fan, Zhiwen and Zhu, Siyu and Dai, Zuozhuo and Tan, Feitong and Tan, Ping},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={2495--2504},
  year={2020}
}

Acknowledgements

Thanks to Xiaoyang Guo for opening source of his excellent works GwcNet and MVSNet-pytorch. Thanks to Yao Yao for opening source of his excellent work MVSNet.