update readme

Author: XiangChen1994

README.md

@@ -17,37 +17,36 @@ Image registration aims to establish spatial correspondence across pairs, or gro
 
 ## Network:<a id="Network"/>
 The kernal idea is to use dep learning network to mimic the process of deforming the template mesh under the guidence of contours.
-![image](https://github.com/XiangChen1994/MR-Net/blob/main/fig/MRNet.png)
+![image](https://github.com/XiangChen1994/DDIR/blob/main/fig/DDIR.png)
 
 ## Repo Contents:<a id="Repo Contents"/>
-This code is based on [Pixel2mesh](https://github.com/nywang16/Pixel2Mesh), where the GCN block and the mesh loss are mainly from it.
-The point feature extraction is partially referred to [PointNet++](https://github.com/charlesq34/pointnet2).
+This code is partially referred to [voxelmorph](https://github.com/voxelmorph/voxelmorph), where the GCN block and the mesh loss are mainly from it.
 
 ## Package dependencies:<a id="Package dependencies"/>
-This repository is based on Python2.7, Tensorflow and Tensorlayer.
-The version of the main packages is as follows,
-- Tensorflow==1.7.0
-- tflearn
+This repository is based on Python3.6, Tensorflow and Keras.
+The versions of the main packages are as follows,
+- Tensorflow==1.5.0
+- Keras==2.2.4
 
 ## Dataset:<a id="Dataset"/>
-Our network is trained based on UKBB dataset. The input contours are extracted from manual segmentation results, and the groun-truth is generated by traditional method (deforming a template mesh conditioned by the corresponding contours). If you want to train MR-Net by yourself but have no access to the UKBB, ACDC dataset could be another option.
+Our network is trained and tested based on UKBB cardiac MR images. If you want to train DDIR by yourself but have no access to the UKBB, ACDC dataset could be another option.
 
 ## Training:<a id="Training"/>
-Use the following command to train the MR-Net.
+Use the following command to train the DDIR.
 ```sh
- CUDA_VISIBLE_DEVICES=0 python train.py  --data_dir path/to/trainfile/
+python train.py --data_dir path/to/trainfile/  --gpu 0 --model_dir path/to/model file/
 ```
 
 ## Testing:<a id="Testing"/>
-Use the following command to test the MR-Net. Chamfer Distance (CD), Earth Mover Distance (EMD), Hausdorff Distance (HD) and Point cloud to point cloud (PC-to-PC) error are evaluated in this paper.
+Use the following command to test the DDIR. Dice score, Hausdorff Distance (HD) and several clinical indices are evaluated in this paper.
 ```sh
-CUDA_VISIBLE_DEVICES=0 python test.py --data_dir path/to/testfile/
+python test.py 0 path/to/model file/  modelname
 ```
 
 ## Demo:<a id="Demo"/>
 To reconstruct 3D cardiac mesh with pretrained model from contours.
 ```sh
-CUDA_VISIBLE_DEVICES=0 python demo.py --test_file demo/test.vtk
+python demo.py 0 path/to/model file/  modelname
 ```
 
 ## Citation:<a id="Citation"/>