# Deep-ADMM-Net

**Repository Path**: cmystal_space/Deep-ADMM-Net

## Basic Information

- **Project Name**: Deep-ADMM-Net
- **Description**: No description available
- **Primary Language**: Matlab
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2019-03-22
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Deep-ADMM-Net

***********************************************************************************************************

This is a testing and training code for Deep ADMM-Net in "Deep ADMM-Net for Compressive Sensing MRI" (NIPS 2016)
 
If you use this code, please cite our paper:

[[1] Yan Yang, Jian Sun, Huibin Li, Zongben Xu. Deep ADMM-Net for Compressive Sensing MRI, NIPS(2016).](http://gr.xjtu.edu.cn/web/jiansun/publications])

http://gr.xjtu.edu.cn/web/jiansun/publications

All rights are reserved by the authors.

Yan Yang -2017/04/05. For more detail, feel free to contact: yangyan92@stu.xjtu.edu.cn


***********************************************************************************************************



## Usage:

1. For testing the trained network 

	1). Load trained network with different stages in main_ADMM_Net_test.m.<br>
	If you apply ADMM-Net to  reconstruct  other MR images, it is best to re-train the models.

	 	The models in './net/network_20' are trained from 100 real MR trainging images with 20% sampling rate.
		The models in './net/network_30' are trained from 100 real MR trainging images with 30% sampling rate.
 

	2). Load  sampling pattern with different sampling ratios in main_ADMM_Net_test.m

   		The mask in './mask/mask_20' is a pseudo radial sampling pattern with 20% sampling rate.
   
	3). Load test image  in main_ADMM_Net_test.m

   		The images in './data/Brain_data' are real-valued brain MR images.
   		The images in './data/Chest_data' are 50 real-valued chest MR testing images in our paper.

	4). Network setting is in  'config.m '.

	5). To test our ADMM-Net, run 'main_ADMM_Net_test.m'


2. For training the networks

	1). The training chest dataset is in './data/ChestTrain_20'.<br>
    	    Run 'Gen_traindata.m' to generate training data, and load  corresponding sampling pattern in this operation. 

	2). Modify the network setting and trainging setting in  'config.m '.

	3). To train ADMM-Net by L-BFGS algorithm, run 'main_netTrain.m' . 

	4). After training, the trained network and the training error are saved in './Train_output'.



***********************************************************************************************************

The testing result of the demo images.

	1) Brain_data1.(20% sampling rate)

	|--------------|  re_LOss  |  re_PSnr  | 
	|--------------|-----------|-----------| 
	|  net-stage7- |  0.0578   |  35.60    |  
	|  net-stage14 |  0.0562   |  35.83    |  
	|  net-stage15 |  0.0561   |  35.85    |  


	2) Brain_data2.(20% sampling rate)

	|--------------|  re_LOss  |  re_PSnr  | 
	|--------------|-----------|-----------|  
	|  net-stage7- |  0.0957   |  30.40    |  
	|  net-stage14 |  0.0929   |  30.65    |  
	|  net-stage15 |  0.0927   |  30.67    |