junyuchen245/ViT-V-Net_for_3D_Image_Registration_Pytorch

How do you setting your data in order to run python train.py?

tphankr opened this issue · 2 comments

Please help me setting the data in order to train. Or help me have the simply data to train ( I also download your datas that mentioned but I can not use it)
image
I assumer use the BraTS_2018 dataset. It looks like:
image
image

Val:
image
image

When I set the above path, the following error occurs:
image

Thank you, very much.

Hi,

The script assumes the images are preprocessed (e.g., intensity normalized into a range of [0,1], skull stripped, and affine registered) and in .pkl format. A single .pkl file consists of one input image and one ground truth. You can find out details about this in "datasets.py".

For Brats, you can follow the preprocessing steps here. However, they normalized images to z-scores. I haven't tested if the model can work with negative values, so it is better to normalize the images into [0,1].

For other public MRI datasets, you will need to do the preprocessing yourself by using FreeSurfer. I've listed some of the commands I used here.

Thanks,
Junyu

I am following your advise. I try to do succesully this task. thank you.