junjun-jiang / rlpa

Hyperspectral Image Classification in the Presence of Noisy Labels (IEEE TGRS, 2019)

The code is for the work:

@article{jiang2019hyperspectral,
  title={Hyperspectral Image Classification in the Presence of Noisy Labels},
  author={Jiang, Junjun and Ma, Jiayi and Wang, Zheng and Chen, Chen and Liu, Xianming},
  journal={IEEE Transactions on Geoscience and Remote Sensing},
  volume={57},
  number={2},
  pages={851--865},
  year={2019}
}

@article{jiang2020learning,
 author={J. {Jiang} and J. {Ma} and X. {Liu}}, 
 journal={IEEE Transactions on Neural Networks and Learning Systems}, 
 title={Multilayer Spectral-Spatial Graphs for Label Noisy Robust Hyperspectral Image Classification}, 
 year={2020}, 
 volume={}, 
 number={}, 
 DOI={10.1109/TNNLS.2020.3029523},}

To generate the file of *_randp.mat for other database, you can use the randpTest.m to generate.

If you need another two datasets (PaviaU and Salinas), please feel free to contact me. Or you can download them from http://www.ehu.eus/ccwintco/index.php/Hyperspectral_Remote_Sensing_Scenes

PaviaU: http://www.ehu.eus/ccwintco/uploads/e/ee/PaviaU.mat, http://www.ehu.eus/ccwintco/uploads/5/50/PaviaU_gt.mat

Salinas: http://www.ehu.eus/ccwintco/uploads/a/a3/Salinas_corrected.mat, http://www.ehu.eus/ccwintco/uploads/f/fa/Salinas_gt.mat

rlpa's Issues

Can it apply to other HSI data set such as KSC?

Hi,

I am currently working on HSI classification with noisy labels. And I think your method is really useful and I just wonder if your method can be applied to more HSI data sets, such as KSC. Since when I look in your code, I wonder how the variable 'Ratio' gets its value. Also, if it can be applied to KSC data set, what value of 'Ratio' is applicable for better classification result?

I appreciate your time and look forward to your response.

Thanks,
Yimin Xu