Skip to content

Code of P2Sharpen: A progressive pansharpening network with deep spectral transformation.

License

Notifications You must be signed in to change notification settings

Baixuzx7/P2Sharpen

Repository files navigation

P2Sharpen

Code of P2Sharpen: A progressive pansharpening network with deep spectral transformation.

Running environment :

python=3.8, pytorch-gpu=1.7.1, matlab = 2018a.

Preparation:

  • Construct the train, validation, test dataset according to the Wald protocol.
  • Put the all the dataset in the root directory, namely TrainFolder, ValidFolder and TestFolder.
  • In each directory, there are four subdirectories, namely pan_label/ ms_label/ pan/ ms/
  • The images in each directory should correspond to each other.

To train :

  • The whole training process contains two part, STNet and P2Net.
  • Please run "transfertrain.py" to learn the spectral tranformation network(STNet).
  • TNet guides the optimization of P2Net, so ensuring the accuracy before the next step.
  • Please run "fusiontrain.py" to learn the progressive pansharpening network (P2Net).

To valid :

  • Use the functions in the file ".\Eval.py" or others to evalute the performance on valid dataset.
  • Pick out the best parameters and save it in path "./Model/P2Net/fusion.pth".

To test :

  • Run the "fusionpredict.py" to generate the pansharpening results.
  • Open the Matlab and run the file ".\Evalution\FusionEval.m".

About

Code of P2Sharpen: A progressive pansharpening network with deep spectral transformation.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published