question about 'affine.npy'

[Shawn0099]

Hi,thanks for your work.
I'm running your code but cannot get good result.So far i've run 50 epochs and get 0.72 dice coefficient.It seems the number is not going higher for many epochs.
There are two differences:1.my dataset is brats 2018 which incliude 285 cases.2.I didn't apply flip or distort.In your paper you mentioned using these data augmentation strategy,but the file'affine.npy'is missing.It would be really helpful if you can share how you do this.

[woodywff]

Thanks for your interest.
The affine.npy will be generated for the first time on the fly when you use /unet3d/data.py's write_image_data_to_file() to write the .h5 file.
If you just need this file, you could also get it from any mri image by np.save('affine', nib.load('*.nii.gz').affine).