Adding 2D Equivariant Feature Extraction Example #8233
Conversation
Signed-off-by: Danny Joca <[email protected]> Signed-off by: Jian Wang <[email protected]> Signed-off by: Ali Gholipour <[email protected]>
Signed-off-by: Danny Joca <[email protected]>
bring local branch up to date
for more information, see https://pre-commit.ci
| @@ -61,3 +61,4 @@ pyamg>=5.0.0 | |||
| git+https://github.com/facebookresearch/segment-anything.git@6fdee8f2727f4506cfbbe553e23b895e27956588 | |||
| onnx_graphsurgeon | |||
| polygraphy | |||
| e3nn | |||
There was a problem hiding this comment.
Hi @djoca77, thanks for your contribution.
I didn't quite understand the purpose of this PR. If the goal is to include a 2D equivariant feature extraction feature, it would be better to integrate it into the core and add relevant test cases, rather than just introducing the test case and dependency here. You can also add some example tutorials in the tutorial repo.
There was a problem hiding this comment.
Hi, thank you for the feedback. Could you elaborate more on what core I should integrate the feature into in the monai folder? Any guidance would be much appreciated.
There was a problem hiding this comment.
Hi @djoca77, you may need to minimize code repetition as much as possible by organizing related components into appropriate structures. For instance, place convolution operations within the network block, and replace loading .nii files withLoadImage from MONAI. If you find no highly generic components to extract, consider adding this as an example in the tutorial for clarity and reference. Thanks.
There was a problem hiding this comment.
Hi @KumoLiu thank you for the help. I think for now I'd like to add this as an example of the equivariant feature extraction instead of a fully fledged functionality. I can make some changes to use some of the MONAI generic functions. Could you give me some pointers on contributing this as an example. I wonder whether I should just add this to the tutorial as you suggested. Thanks.
Fixes # .
Description
A few sentences describing the changes proposed in this pull request.
Adds an example of a 2D model that extracts equivariant features from a 2D slice of a 3D brain MRI volume. Changes also include adding an optional dependency and a testing data that is currently in the testing_data folder but will be added to the MONAI-extra-test-data release. Project-MONAI/MONAI-extra-test-data#17
Types of changes
./runtests.sh -f -u --net --coverage../runtests.sh --quick --unittests --disttests.make htmlcommand in thedocs/folder.