2D Tensor Support

The transforms in your package are incredibly useful and unique, but for my project I need to apply them on 2D images.

I can likely develop a work-around using the 3D transforms, however I saw on issue #22 and issue #53 that you have previously discussed implementing 2D image support and have since closed these issues. Despite these issues being marked resolved, when using the transforms with the tensor support method (as outlined in #139 ), I get the following error:

RuntimeError: The input tensor must have 4 dimensions (channels, i, j, k), but has 3: torch.Size([1, 120, 384])

This was from trying to feed a tensor with shape (channels, i, j). This error makes it seem like the transforms are still only for 3D inputs? I have tried this with a variety of the transforms and get the same results.

Is there a method I’m missing for applying your transforms with 2D images? Thanks so much in advance for any help you can offer.

[jillianlee]

These all work perfectly for me! Thank you so much!

I don’t need any more transforms for 2D, but thank you though. This helps a lot, I really appreciate your work!

[Chao86]

@romainVala has already replied to you in #643 (comment), please don’t post the same here, in a closed issue. If you find something is not working as you expected, you can open a new issue and follow the instructions: https://github.com/fepegar/torchio/issues/new?assignees=&labels=&template=not_working.md&title=

Sorry , I will refresh to browse the coomment https://github.com/fepegar/torchio/discussions/643#discussioncomment-1270113. Thank you for reminding me!