The code is a pytorch implementation of the paper: A Normalized Disparity Loss for Stereo Matching Networks, IEEE Robotics and Automation Letters, 2022.
-
Requirements: PyTorch>=1.0.0, Python3, tensorboardX
-
Train:
- Train PSMNet on SceneFlow
''' sh tool/run.sh SF PSMNet '''
Adjust the "xx.py" to "train_PSMNet.py" in run.sh and the paths or parameters in ./config/SF/SF_PSMNet.yaml
- Train PSMNet with normalized loss on SceneFlow
''' sh tool/run.sh SF PSMNet_normloss '''
Adjust the "xx.py" to "train_PSMNet_normloss.py" in run.sh and the paths or parameters in ./config/SF/SF_PSMNet_normloss.yaml
- For the training on other datasets, the procedure is similar.
-
Test:
- Test the trained model:
''' sh tool/run.sh SF PSMNet_normloss '''
Adjust the "xx.py" in run.sh to "test_PSMNet_normloss.py" and the paths or parameters in ./config/SF/PSMNet_normloss.yaml
-
Visualization: tensorboard --logdir=xx --port=6789
If you find this code useful in your research, please cite:
@article{Chen2022,
author = {Shuya Chen and Zhiyu Xiang and Peng Xu and Xijun Zhao},
title = {A Normalized Disparity Loss for Stereo Matching Networks},
journal = {IEEE Robotics and Automation Letters},
year = {2022}
}
The code is partly based on the PSMNet and PSPNet. Thanks to these excellent work.
@inproceedings{chang2018pyramid,
title={Pyramid Stereo Matching Network},
author={Chang, Jia-Ren and Chen, Yong-Sheng},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
pages={5410--5418},
year={2018}
}
@inproceedings{zhao2017pspnet,
title={Pyramid Scene Parsing Network},
author={Zhao, Hengshuang and Shi, Jianping and Qi, Xiaojuan and Wang, Xiaogang and Jia, Jiaya},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
year={2017}
}