Awesome

RSSAN-Hyperspectral-Image

Residual Spectral–Spatial Attention Network for Hyperspectral Image Classification 2020 IEEE TGRS

10.1109/TGRS.2020.2994057

https://ieeexplore.ieee.org/abstract/document/9103247

论文代码复现

The pytorch framework used in this code. Instead of the TensorFlow framework used in the article, modify it yourself if necessary.

本代码采用的pytorch框架。而不是文章采用的TensorFlow框架，如有需要自行修改。
RSSAN-Hyperspectral-Image
--Dateset
--function
--model
--resulit
--main.py
--README.md
--train.log

environment：
python 3.8.5
numpy 1.19.2
scikit-learn 0.23.2
tensorflow 2.5.0
torch 1.9.0
....

The running results are saved in the result folder.

运行结果均保存在result文件夹中。

运行方法
python main.py

Parameter setting:
参数设置：

epoch: 200
patch_size: 17
train batch_size: 16
test batch_size: 16 lr: IN,PU 0.0003 KSC 0.0001
optimizer:RMSprop
depth:PU 8 IN,KSC 32
kernel_size: 3

PU：OA, AA, kappa: [0.9916494087781415, 0.9869853556897955, 0.9889365386137344]
each_acc [98.64, 99.77, 95.03, 99.44, 99.79, 99.89, 97.85, 98.33, 99.55]

KSC：OA, AA, kappa: [0.8999450247388675, 0.8738564837073592, 0.8883809689177758]
each_acc [99.06, 64.5, 97.19, 57.14, 85.71, 94.38, 100.0, 99.67, 100.0, 64.18, 100.0, 74.64, 99.54]

IN：OA, AA,kappa: [0.9861789752896831, 0.9721097015738396, 0.9842439139710597]
each_acc [100.0, 98.4, 98.97, 93.94, 100.0, 99.41, 94.74, 100.0, 78.57, 97.5, 98.49, 98.79, 100.0, 99.55, 97.03, 100.0]

问题：
KSC数据集的表现很差

2021.8.19

将论文中的消融实验加上了，自行运行，对比结果。