macrohuang1993/ESNet

ESNet: An Efficient Stereo Matching Network

Code in PyTorch for paper "ES-Net: An Efficient Stereo Matching Network" submitted to IROS 2021 [Paper Link].

Dependency

Python 3.6

PyTorch(1.2.0+)

torchvision 0.2.0

Usage

"networks/ESNet.py" and "networks/ESNet_M.py" contains the implementation of the proposed efficient stereo matching network.

To train the ESNet on SceneFlow dataset, you will need to modify the path to the dataset in the "exp_configs/esnet_sceneflow.conf". Then run

dnn=esnet_sceneflow bash train.sh

Citation

If you find the code and paper is useful in your work, please cite

@misc{huang2021esnet,
    title={ES-Net: An Efficient Stereo Matching Network},
    author={Zhengyu Huang and Theodore B. Norris and Panqu Wang},
    year={2021},
    eprint={2103.03922},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}