This the python (Theano) implementation of the testing phase for MONet, a CNN for SAR despeckling described in the paper Multi-Objective CNN Based Algorithm for SAR Despeckling.
For the pytorch versions go to Pytorch-MONet
MONet is 17 layers CNN with skip connection and a multi-objective cost function L. L is composed of three terms: MSE between output and reference, Kullback-Leibler divergence between estimated noise distribution and the theoretical one, and an edge loss computed on output and the reference. The architecture is shown in the following
This a an example of the results on data with simulated speckle
| Noisy Image | Noise-Free Reference | MONet - Output |
|---|---|---|
 |
 |
 |
Sergio Vitale (contact person, sergio.vitale@uniparthenope.it); Giampaolo Ferraioli (giampaolo.ferraioli@uniparthenope.it); Vito Pascazio (vito.pascazio@uniparthenope.it)
Copyright (c) 2020 Dipartimento di Ingegneria and Dipartimento di Scienze e Tecnologie of Università degli Studi di Napoli "Parthenope".
All rights reserved. This work should only be used for nonprofit purposes.
By downloading and/or using any of these files, you implicitly agree to all the terms of the license, as specified in the document LICENSE.txt (included in this directory)
This code is written on Ubuntu system for Python2.7 and uses Theano library.
The command to install the requirements is:
cat requirements.txt | xargs -n 1 -L 1 pip2 install
If you use a python editor
- install Anaconda
- install requirements and spyder editor with conda
- edit main.py and run
- cuda = 8
- cudnn = 5
-
imgs folder contains three samples images with simulated single look speckle in amplitude format; the sample image are taken from UC Merced LandUse Dataset. Three differente can be tested:
- baseballdiamond
- golfcourse
- storagetanks
-
model folder contains the pre-trained network
-
run test without GPU
python main.py -a <AREA>
- run with GPU
PATH=<CUDAPATH>:$PATH python main.py -g -a <AREA>