The shape of the output is not the same as the original shape of my data--UNETR

[daisysong1]

I checked that the input in the validation was already a miniaturized shape.
I wanna know if this has anything to do with transform.
What changes should I make if I want to end up with the original shape ?
If anyone can help me, thank you very much!

[KumoLiu]

Hi @daisysong1, did you follow this tutorial? https://github.com/Project-MONAI/tutorials/blob/main/3d_segmentation/unetr_btcv_segmentation_3d.ipynb
If so, removing Spacingd and CropForegroundd can help you keep data after val_transforms with the original shape.

BTW, if you remove Spacingd and CropForegroundd in the validation, also update the train_transforms to avoid performance drop due to differences between the training and the validation dataset.

Hope it can help you, thanks!

[daisysong1]

Thank you very much for your reply. I'll train again and see how accurate it is.