This is the reserch code of the IEEE Transactions on Image Processing 2022 paper.
The details of files and directories in this repository are shown as follows.
Item | Description |
---|---|
./Data | The directory containing hyperspectral images. |
./result_RIAN | The directory containing hyperspectral images. |
attention.py | Code of the proposed CSpeA and RSpaA modules. |
HSI_Data_Preparation_Houston.py | Code for splitting Houston2013 data set into training set and testing set. |
HSI_Data_Preparation_PU.py | Code for splitting Pavia University data set into training set and testing set. |
HSI_Data_Preparation_Salinas.py | Code for splitting Salinas data set into training set and testing set. |
RIAN.py | The main manuscript of the proposed RIAN. |
utils_houston.py | Parameter settings for Houston2013 data set. |
utils_PU.py | Parameter settings for Pavia University data set. |
utils_salinas.py | Parameter settings for Salinas data set. |
Requirements:
Python 2.7
tensorflow-1.4.1
scikit-image
1) Split training set and testing set.
Three data sets are used: Houston_2013,Pavia University and Salinas, which can be downloaded from Google Drive https://drive.google.com/file/d/1BkwgTRh3JtKqQcLQ7DkMv98jChs15JmB/view?usp=sharing.
Run "HSI_Data_Preparation_Houston.py", "HSI_Data_Preparation_PU.py" and "HSI_Data_Preparation_Salinas.py " to get training set and testing set of each data set.
2) Run "python RIAN.py" for training and testing.
Revise lines 27-33 to change data set. Testing results are saved in the directory "result_RIAN".
3) Run "result_RIAN/DATASETNAME/acc_in_testing_set.m" in Matlab to calculate evaluation metrics such as OA, AA and kappa.
"DATASETNAME" is the name of data set.
4) For the codes of compared methods in the paper, please refer to:
DHCNet: https://github.com/ordinarycore/DHCNet
SSRN: https://github.com/zilongzhong/SSRN
1-D, 2D-CNN:https://github.com/nshaud/DeepHyperX
SMBN, SFFN:http://www.escience.cn/people/LeyuanFang/index.html
MGCN: https://github.com/danfenghong/IEEE_TGRS_GCN
If you find the code and dataset useful in your research, please consider citing our paper.
@ARTICLE{9785505,
author={Zheng, Xiangtao and Sun, Hao and Lu, Xiaoqiang and Xie, Wei},
journal={IEEE Transactions on Image Processing},
title={Rotation-Invariant Attention Network for Hyperspectral Image Classification},
year={2022},
volume={31},
doi={10.1109/TIP.2022.3177322}
}