NeRT: Implicit Neural Representations for General Unsupervised Turbulence Mitigation

Weiyun Jiang, Yuhao Liu, Vivek Boominathan, Ashok Veeraraghavan

Project Page

Set up environment

conda env create -f implicit_turbu.yml
conda activate implicit_turbu

Dataset

• Put the desired dataset under data/Ablation_data/
• Download the precomputed p2s model parameters from Google Drive
• Unzip p2s_data.zip under data/p2s_data

High-Level structure

The code is organized as follows:

• networks/ contains some basic neural network building blocks.
• utils/ contains utility functions, most promintently related to reading images.
• NeRT.ipynb contains demo codes.

Reproducing experiments

Run the NeRT.ipynb using jupyter notebook

Citation

@article{jiang2023nert_general,
    title={Nert: Implicit neural representations for general unsupervised 
           turbulence mitigation},
    author={Jiang, Weiyun and Liu, Yuhao and Boominathan, Vivek and 
            Veeraraghavan, Ashok},
    journal={arXiv preprint arXiv:2308.00622},
    year={2023}
}

@inproceedings{jiang2023nert,
    title={Nert: Implicit neural representations for unsupervised atmospheric 
           turbulence mitigation},
    author={Jiang, Weiyun and Boominathan, Vivek and Veeraraghavan, Ashok},
    booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and 
               Pattern Recognition Workshops},
    pages={4235--4242},
    year={2023}
}