Documentation of spatial shape constraints for the networks

[wyli]

Is your feature request related to a problem? Please describe.
follow up of discussion #6641

would be great to add input spatial size vs model configuration analysis for all the networks such as
swinunetr and dints.

the form of analysis could be detailed tutorials (example https://github.com/Project-MONAI/tutorials/blob/main/modules/UNet_input_size_constrains.ipynb) and/or actual code logic to check the compatibility of the input shape and model configuration.

[nhhung1810]

Hi @wyli , I'd like to work on this issue, can you suggest me some related resources that you think is helpful?

[wyli]

sure, that's great @nhhung1810, please see:

• how to prepare a pull request https://github.com/Project-MONAI/MONAI/blob/dev/CONTRIBUTING.md
• would be great to revise the related docstring with a bit more details, following this example

  MONAI/monai/networks/nets/unet.py

  Lines 32 to 54 in 6f9cf6b

  	Enhanced version of UNet which has residual units implemented with the ResidualUnit class.
  	The residual part uses a convolution to change the input dimensions to match the output dimensions
  	if this is necessary but will use nn.Identity if not.
  	Refer to: https://link.springer.com/chapter/10.1007/978-3-030-12029-0_40.
  	Each layer of the network has a encode and decode path with a skip connection between them. Data in the encode path
  	is downsampled using strided convolutions (if `strides` is given values greater than 1) and in the decode path
  	upsampled using strided transpose convolutions. These down or up sampling operations occur at the beginning of each
  	block rather than afterwards as is typical in UNet implementations.
  	To further explain this consider the first example network given below. This network has 3 layers with strides
  	of 2 for each of the middle layers (the last layer is the bottom connection which does not down/up sample). Input
  	data to this network is immediately reduced in the spatial dimensions by a factor of 2 by the first convolution of
  	the residual unit defining the first layer of the encode part. The last layer of the decode part will upsample its
  	input (data from the previous layer concatenated with data from the skip connection) in the first convolution. this
  	ensures the final output of the network has the same shape as the input.
  	Padding values for the convolutions are chosen to ensure output sizes are even divisors/multiples of the input
  	sizes if the `strides` value for a layer is a factor of the input sizes. A typical case is to use `strides` values
  	of 2 and inputs that are multiples of powers of 2. An input can thus be downsampled evenly however many times its
  	dimensions can be divided by 2, so for the example network inputs would have to have dimensions that are multiples
  	of 4. In the second example network given below the input to the bottom layer will have shape (1, 64, 15, 15) for
  	an input of shape (1, 1, 240, 240) demonstrating the input being reduced in size spatially by 2**4.

• the docstring format https://www.sphinx-doc.org/en/master/usage/restructuredtext/basics.html
• the docstring style https://google.github.io/styleguide/pyguide.html#38-comments-and-docstrings

[nhhung1810]

all the networks such as swinunetr and dints

Hi @wyli , I think that you refer to all networks under monai/networks/nets? If such, can you suggest some high priority networks first?

Also, if the number of networks need to be done is a lot, I'll split up into small PRs for each networks, do you fine with that?

[wyli]

Yes @nhhung1810, separate PRs would be great, high priority ones are

• https://github.com/Project-MONAI/MONAI/blob/dev/monai/networks/nets/basic_unetplusplus.py
• https://github.com/Project-MONAI/MONAI/blob/dev/monai/networks/nets/segresnet_ds.py
• https://github.com/Project-MONAI/MONAI/blob/dev/monai/networks/nets/swin_unetr.py
• https://github.com/Project-MONAI/MONAI/blob/dev/monai/networks/nets/dints.py

please feel free to start with simple PRs.. we can use this ticket to track the progress (cc @vikashg for any further comments)

[nhhung1810]

Sure, I will start with UNET++ first

[nhhung1810]

Hi @wyli I have create a PR here Project-MONAI/tutorials#1476, mentioned in the above Issue Project-MONAI/tutorials#1475.

It's not completed, I'm definitely need more advise on this

[vikashg]

Good job @nhhung1810

[vikashg]

@nhhung1810 is it completed?

[nhhung1810]

Oh it's complete @vikashg, it's in the PR above, for 1 model (UNET++) only. I haven't worked on this since then @vikashg