Rand3DElasticDict

[lyuxiaobai]

For segmentation tasks, are there any suggestions for the values ​​of sigma and magnitude_range of Rand3DElasticDict? If I pass the image and real annotations into this method at the same time, how should the mode be written? I want to use bilinear for images and nearest for real labeling, but I don’t know how to achieve this

[lyuxiaobai]

[lyuxiaobai]

train_transforms = Compose([
LoadNiftid(keys={'image','label'}),

AddChanneld(keys={'image','label'}),

ScaleIntensityd(keys={'image'}),
Rand3DElasticDict(keys={'image','label'},sigma_range=(15,25),magnitude_range=(0,5),prob=0.4,mode='bilinear,nearest'),
RandFlipDict(keys={'image','label'}),
ToTensor()

])
代码如上

[Nic-Ma]

Hi @lyuxiaobai ,

Thanks for your interest in MONAI.
I saw you already set mode='bilinear, nearest' in your code for image and label.
So could you please share more details about your question?

Thanks.