ATLAS_100%group_link

**ATLAS dataset using 170 samples(100%) to do CarveMix(In fact, out of 170 samples, validation sets need to be removed, so only 136 participate in CarveMix ) 
to generate a total of 1,000 samples(contain 170 training and validation) for nnUNet architecture.**
All data related to this case, nnUNet training model and analysis of predicted results are available at the link 
'''
https://pan.baidu.com/s/15sJeRLIHWjx1-sjB32lKmg
code:ucue
'''
From the files in the link, you can see how a total of 220 ATLAS datasets are divided into imagesTr and imagesTs in nnUNet. It is convenient for you to reproduce the code.
Because it is the initial version of "CarveMix", the generated sample does not contain the keyword CarveMix, but is distinguished by serial number 0-829.