padim arguments improvements

[jpcbertoldo]

Description

Fix #662 by creating the parameter n_features (nb of features/axes retained by the dimension reduction).

I also made it possible to use a custom backbone without modifying the source code.
This was limited by this hardcoded dict

If the backbone was not in the dict then the orig_dims and emb_scale could not be set.
These two don't need to be hardcoded, I made it be deduced from a dryrun feature extraction.

Changes

• Bug fix (non-breaking change which fixes an issue)
• Refactor (non-breaking change which refactors the code base)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

Checklist

• My code follows the pre-commit style and check guidelines of this project.
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing tests pass locally with my changes

[samet-akcay]

Thanks @jpcbertoldo. The timing of this PR is also great, given that we wanted to add Efficient backbone support.

[jpcbertoldo]

Cool!

I will write some the doc for it.
Does this feature need tests in your opinion?

[ashwinvaidya17]

Looks good to me

[jpcbertoldo]

done
i further simplified a bit

[jpcbertoldo]

Maybe it would be good to add a unit test for this function

@djdameln it didn't feel right to have such a specific test.

Instead i extracted the dryrun part and tested that. The logic in _deduce_dims is trivial now.

[samet-akcay]

There is going to be merge conflicts between this PR and #675. @ashwinvaidya17 and @jpcbertoldo, can you guys sync how to minimise/avoid these conflicts?

[jpcbertoldo]

The conflict is only in anomalib/models/components/feature_extractors/__init__.py right?

Doesnt look so hard, whicever is merged second rebases.

@ashwinvaidya17 could you check if the new function in that file in #675 would work with the torchfx extractor? If not then it needs a subcase.

[ashwinvaidya17]

The conflict is only in anomalib/models/components/feature_extractors/__init__.py right?

Doesnt look so hard, whicever is merged second rebases.

@ashwinvaidya17 could you check if the new function in that file in #675 would work with the torchfx extractor? If not then it needs a subcase.

I am working on a design to make the feature extractor configurable. I think we can move dimension computation into the feature extractor class. Since the design might take time we can merge this one first and then I can update PR #675

[jpcbertoldo]

The conflict is only in anomalib/models/components/feature_extractors/__init__.py right?
Doesnt look so hard, whicever is merged second rebases.
@ashwinvaidya17 could you check if the new function in that file in #675 would work with the torchfx extractor? If not then it needs a subcase.

I am working on a design to make the feature extractor configurable. I think we can move dimension computation into the feature extractor class. Since the design might take time we can merge this one first and then I can update PR #675

Cool : )