README.md

Spectral-Spatial Unified Networks for Hyperspectral Image Classification

Keras implementation of our method for hyperspectral image classification.

Paper

Spectral–Spatial Unified Networks for Hyperspectral Image Classification

Please cite our papers if you find it useful for your research.

@article{ssun,
  author={Y. Xu and L. Zhang and B. Du and F. Zhang},
  journal={IEEE Trans. Geos. Remote Sens.},
  title={Spectral-Spatial Unified Networks for Hyperspectral Image Classification},
  year={2018},
  volume={56},
  number={10},
  pages={5893-5909},
  ISSN={0196-2892},
  month={Oct}
}

@inproceedings{bglstm,
  title={A Band Grouping Based LSTM Algorithm for Hyperspectral Image Classification},
  author={Y. Xu and B. Du and L. Zhang and F. Zhang},
  booktitle={CCF Chinese Conference on Computer Vision},
  pages={421--432},
  year={2017},
  organization={Springer}
}

Installation

• Install Keras 2.2.4 from https://github.com/keras-team/keras with Python 3.6.

  • Note: This repo is now updated with the Tensorflow backend engine. We have tested the code with Tensorflow 1.13. For the Theano backend users, please refer to https://keras.io/#configuring-your-keras-backend for technical support.

• Clone this repo.

git clone https://github.com/YonghaoXu/SSUN

Dataset

• Download the Pavia University image and the corresponding annotations.

Usage

• Replace the file path for the hyperspectral data in HyperFunctions.py with yours.
• Run SSUN.py.
• Change the s1s2 index in SSUN.py to switch from different grouping strategies.
  • Left: Strategy 1 s1s2 = 1
  • Right: Strategy 2 s1s2 = 2

Note

• 12/2019: Update the code with the Tensorflow backend engine.