readme.md

[ECCV 2024] Region-Aware Sequence-to-Sequence Learning for Hyperspectral Denoising

The official PyTorch implementation of RAS2S.

Install the environment

git clone https://github.com/MIV-XJTU/RAS2S.git
cd RAS2S
conda env create -f environment.yml
conda activate py37

Model

The primary implementation of the RAS2S can be found in the following directory:

basic/models/competing_methods/s2s_backbone

Other competing methods can also be placed in the same folder.

Dataset

Download

The entire ICVL dataset download from here.

And thanks to QRNN3D. The division and processing of dataset in this paper are consistent with theirs.

Processing

The training and testing samples are listed in the following file:

basic/utility/icvl_train_list.txt
basic/utility/icvl_test_list.txt

Prepare the LMDB dataset for training.

cd basic
python utility/lmdb_data.py

Prepare the MAT dataset for testing.

cd basic
python utility/mat_data.py

Training and Evaluation

Configuration modification

Once the training and testing datasets are prepared, please ensure that the relevant data paths trainDir, valDir, and testDir are correctly set in the configuration file options/s2s_hsid.yml.

Training

Ensure that the model checkpoint path save_path is correctly set in the configuration file.

cd basic
python run.py -method s2s -mode train

Testing

Please modify the model weight path resumePath in the configuration file.

cd basic
python run.py -method s2s -mode test

Citation

If any parts of our paper and code help your research, please consider citing us and giving a star to our repository.

@inproceedings{xiao2024region,
  title={Region-Aware Sequence-to-Sequence Learning for Hyperspectral Denoising},
  author={Xiao, Jiahua and Liu, Yang and Wei, Xing},
  booktitle={European Conference on Computer Vision},
  year={2024}
}

Acknowledgements

Todo.