shuangxu96/CSC-MMFN

CSC-MMFN

Deep Convolutional Sparse Coding Networks for Image Fusion. arxiv

Shuang Xu *, Zixiang Zhao *, Yicheng Wang, Kai Sun, Chunxia Zhang, Junmin Liu, Jiangshe Zhang. (* equal contributions)

Requirements

• pytorch (my version is 1.3.0)
• kornia
• tensorboardX
• h5py
• xlwt

Train & Test

Retrain and Test CSC-MMFN

The train and test codes are available lines 7-105 and 111-142 of train.py. If you want to retrain this network, you should:

• Please download and unzip the dataset into the folder MMF_data/scale2. My folder is organized as follows:

    mypath
    ├── train
    │   ├── balloons.mat 
    │   ├── beads.mat
    │   └── ...
    ├── test
    │   ├── real_and_fake_apples.mat
    │   ├── real_and_fake_peppers.mat
    │   └── ...
    ├── validation
    │   ├── paints.mat
    │   ├── photo_and_face.mat
    │   └── ...
    └── ...

• Run lines 7-105 for training.
• Run lines 111-142 for testing.

Test MEFN with Pretrained Weights

A pretrained weight file is provided. If you do not want to retrain this model, please run test.py.

Reference

@article{DBLP:journals/corr/abs-2005-08448,
  author    = {Shuang Xu and
               Zixiang Zhao and
               Yicheng Wang and
               Chunxia Zhang and
               Junmin Liu and
               Jiangshe Zhang},
  title     = {Deep Convolutional Sparse Coding Networks for Image Fusion},
  journal   = {CoRR},
  volume    = {abs/2005.08448},
  year      = {2020},
  url       = {https://arxiv.org/abs/2005.08448},
  archivePrefix = {arXiv},
  eprint    = {2005.08448},
  timestamp = {Fri, 22 May 2020 18:01:19 +0200},
  biburl    = {https://dblp.org/rec/journals/corr/abs-2005-08448.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}