- This is a pytorch implementation of part of 3DGAUnet
- Python >= 3.7.9
- Pytorch >= 1.6.0
- tensorboardX >= 2.1
- matplotlib >= 2.1
Volumetric data in .mat format should be placed in'/src/volumetric_data/', and change the directory in 'paramt.py' accordingly.
cd src
run python main.py
on GPU or CPU. Of course, you need a GPU for training until you get good results. I used one GeForce RTX 4090 in my experiments on 3D models with a resolution of 64x64x64.
- Model weights and some 3D reconstruction images would be logged to the
outputs
folders, respectively, for everymodel_save_step
number of step inparamt.py
. You can play with all parameters inparamt.py
.
To generate volumetric data from the trained model, you can run python main.py --test=True
to call tester.py
.
Pretrained models are in the outputs
folder. Then run python main.py --test=True --model_name=pancreas_pretrained
. You will find the outputs in the test_outputs
folder within pancreas_pretrained
.