The PyTorch implements of RTF: Recursive TransFusion for Multi-Modal Image Synthesis
The overview of our RTF framework.
Our method can synthesis clear and nature images and outperforms other state-of-the-art methods on many datasets.
Experiment results on BraTS2020 dataset.
Experiment results on ISLES2015 dataset.
Experiment results on Taskonomy dataset.
python 3.8.10
pytorch 1.8.1
torchvision 0.9.1
tqdm 4.62.1
numpy 1.20.3
SimpleITK 2.1.0
scikit-learn 0.24.2
opencv-python 4.5.3.56
easydict 1.9
tensorboard 2.5.0
Download the datasets from the official way and rearrange the files to the following structure.
MICCAI_BraTS2020_TrainingData
├── flair
│ ├── BraTS20_Training_001_flair.nii.gz
│ ├── BraTS20_Training_002_flair.nii.gz
│ ├── BraTS20_Training_003_flair.nii.gz
│ ├── ...
├── t2
│ ├── BraTS20_Training_001_t2.nii.gz
│ ├── BraTS20_Training_002_t2.nii.gz
│ ├── BraTS20_Training_003_t2.nii.gz
│ ├── ...
├── t1
│ ├── BraTS20_Training_001_t1.nii.gz
│ ├── BraTS20_Training_002_t1.nii.gz
│ ├── BraTS20_Training_003_t1.nii.gz
│ ├── ...
├── t1ce
│ ├── BraTS20_Training_001_t1ce.nii.gz
│ ├── BraTS20_Training_002_t1ce.nii.gz
│ ├── BraTS20_Training_003_t1ce.nii.gz
│ ├── ...
SISS2015_Training
├── 1
│ ├── VSD.Brain.XX.O.MR_T2.70616
│ ├── VSD.Brain.XX.O.MR_T2.70616.nii
│ ├── VSD.Brain.XX.O.MR_T1.70615
│ ├── VSD.Brain.XX.O.MR_T1.70615.nii
│ ├── VSD.Brain.XX.O.MR_Flair.70614
│ ├── VSD.Brain.XX.O.MR_Flair.70614.nii
│ ├── VSD.Brain.XX.O.MR_DWI.70613
│ ├── VSD.Brain.XX.O.MR_DWI.70613.nii
├── 2
│ ├── VSD.Brain.XX.O.MR_T2.70622
│ ├── VSD.Brain.XX.O.MR_T2.70622.nii
│ ├── VSD.Brain.XX.O.MR_T1.70621
│ ├── VSD.Brain.XX.O.MR_T1.70621.nii
│ ├── VSD.Brain.XX.O.MR_Flair.70620
│ ├── VSD.Brain.XX.O.MR_Flair.70620.nii
│ ├── VSD.Brain.XX.O.MR_DWI.70619
│ ├── VSD.Brain.XX.O.MR_DWI.70619.nii
├── 3
│ ├── ...
We use the sample dataset of Taskonomy, where a link to download the sample dataset is here. You can see CEN code for detailed instructions and tips.
Modify the corresponding dataset path for training configuration and run the following command to train.
python train.py
Modify the corresponding dataset path for testing configuration and run the following command to test.
python test.py