How to load multi-modality data and multi-class mask easily?

[Bigsealion]

Hi,
For example, I have 2 modality data with the same shape, and the mask is multi-class. I want to augment this data and input it into a PyTorch network. And I want to stack the 2 modalities as 2 channels, and set the mask as one-hot (means the output of the network will have the same channels with class).

Do I have to stack the data with dim-1 manually after augment? And do I have to apply one-hot to mask manually? Is there any default method in Torchio that can do this?

Thanks!

[romainVala]

Hi
did you see the different tutorial
In this one, you get the answer with the SubjectsDataset
general tutorial

and it is worth to have a look at the other ones found here
https://github.com/fepegar/torchio/tree/main/tutorials

[fepegar]

Additionally, please take a look at OneHot.

[Bigsealion]

Thanks for your reply!
After reading some discussions and tutorials, I have found some solutions.

For loading multi-modality data (in same geomerty) to channels , using img=tio.ScalarImage([img1_path, img2_path]) is easily.
And OneHot is also useful.