Skip to content
FBPConvNet for computed tomography
Branch: master
Clone or download
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
matconvnet-1.0-beta23 fetch Jun 21, 2017
pretrain pretrain Jun 21, 2017
.gitignore remove mat file Jun 21, 2017
README.md Update README.md Nov 6, 2017
cnn_fbpconvnet.m fetch Jun 21, 2017
cnn_train_fbpconvnet.m Update cnn_train_fbpconvnet.m Jul 26, 2017
cnn_unet_init.m erase comments Jun 21, 2017
computeRegressedSNR.m fetch Jun 21, 2017
dev_fun.m fetch Jun 21, 2017
evaluation.m Update evaluation.m Jun 21, 2017
main.m fetch Jun 21, 2017
vl_euclideanloss.m fetch Jun 21, 2017
vl_simplenn_fbpconvnet.m fetch Jun 21, 2017
vl_simplenn_fbpconvnet_eval.m Update vl_simplenn_fbpconvnet_eval.m Apr 10, 2018

README.md

FBPConvNet - Matlab

Deep Convolutional Neural Network for Inverse Problems in Imaging
http://ieeexplore.ieee.org/document/7949028/

Readme

  1. Before launching FBPConvNet, the MatConvNet (http://www.vlfeat.org/matconvnet/) have to be properly installed. (For the GPU, it needs a different compilation.)
  2. Properly modify matconvnet path in main.m and evaluation.m files.
  3. To start, download 2 links;
    (1) pretrained network : https://drive.google.com/open?id=0B9fSVcoxJuVwMVJ1eWFPdEEwbWs , then put this network into 'pretrain' folder
    (2) dataset : https://drive.google.com/open?id=0B9fSVcoxJuVwMDlxbXdvcTRaM2M just place this data in the same folder with main.m
  4. Use main.m for training. After training, run evaluation.m for deploy of test data set.

*note : phantom data set (x20) is only provided. SNR value may be slightly different with our paper.
*note : these codes mainly ran on Matlab 2016a with GPU TITAN X (architecture : Maxwell)
contact : Kyong Jin (kyonghwan.jin@gmail.com),

special thanks to Junhong Min (Senior Researcher at Samsung Electronics) for providing initial codes.

You can’t perform that action at this time.