/3DOC-SSAN-PyTorch

A PyTorch implementation of paper: Hyperspectral Image Classification Based on 3-D Octave Convolution With Spatial–Spectral Attention Network (https://ieeexplore.ieee.org/document/9140029)

Primary LanguagePythonMIT LicenseMIT

3DOC-SSAN-PyTorch

Unofficial PyTorch implementation of paper Hyperspectral Image Classification Based on 3-D Octave Convolution With Spatial–Spectral Attention Network

Requirements

To install requirements

conda env create -f environment.yml

Preprocess dataset

Simple example on Pavia University dataset

  • Download Pavia University dataset and move it to ./datasets/Pavia_University/
  • Run data processing
python ./datasets/data_process_pavia_university.py

Training

python ./train.py

Result

Result in Pavia University Dataset

Metric Reported Our Implementation
OA 99.87 ± 0.08% 99.81%
AA 99.85 ± 0.15% 99.82%
Kappa 99.82 ± 0.16% 99.74%

Acknowledgements

3DOC-SSAN-TensorFlow-Official