ENH: Update connected component testing baselines

[CavRiley]

Description

This PR updates the baseline images for ConnectedComponent tests following the changes in PR #5622, which replaced vnl_sample.h with <random>. The switch to the standard library random number generator produced different random sequences, causing the existing baseline images to fail validation.

Related Issues/PRs

Follow-up to #5622

PR Checklist

• No API changes were made (or the changes have been approved)
• No major design changes were made (or the changes have been approved)
• Added test (or behavior not changed)
• Updated API documentation (or API not changed)
• Added license to new files (if any)
• Added Python wrapping to new files (if any) as described in ITK Software Guide Section 9.5
• Added ITK examples for all new major features (if any)

Testing

Baseline images were regenerated by running the ConnectedComponent tests with the updated random number generator. All tests now pass with the new baselines.

[github-actions]

Thank you for contributing a pull request! 🙏

Welcome to the ITK community! 🤗👋☀️

We are glad you are here and appreciate your contribution. Please keep in mind our community participation guidelines. 📜
More support and guidance on the contribution process can be found in our contributing guide. 📖

This is an automatic message. Allow for time for the ITK community to be able to read the pull request and comment
on it.

[N-Dekker]

@CavRiley Would it be nice to squash your commit and mine together? I feel that one commit cannot live without the other 😸

If you like that, feel free to squash the two commits, and then please add me as co-author! (At the end of your commit message, Co-authored-by: Niels Dekker <[email protected]>.)

[blowekamp]

What were the .1 alternate bases lines generated?

They have some ### times been different results from different platforms, when the results are not always consistent.

[dzenanz]

Ah, well noticed Brad. The baselines should be replaced, not added.

[CavRiley]

What were the .1 alternate bases lines generated?

They have some ### times been different results from different platforms, when the results are not always consistent.

The .1 alternate baselines were regenerated to account for platform-specific differences in the test outputs.

I generated the primary .png baseline on a Mac machine and the .1 alternative on a Linux machine. During testing, I noticed that the Mac-generated baseline did not fix the test failures on Linux, which is why the .1 alternative was changed.

Let me know if this answered your question!

[N-Dekker]

The baselines should be replaced, not added.

It seems reasonable to me to still keep the old baseline as well, so that at least the final ITK 5.x release can still pass the tests. Right?

[dzenanz]

Yes, that does answer my question Cavan.

Niels, we don't need old test baselines in the main branch. During release process, then-current test data is archived.

[hjmjohnson]

@CavRiley Nice work!

[N-Dekker]

Thanks @CavRiley ! Cool!

Approved, but please consider squashing the two commits 😃

[CavRiley]

Thanks @CavRiley ! Cool!

Approved, but please consider squashing the two commits 😃

Sorry I just got to this. I hope it looks alright!

[N-Dekker]

Thank you for squashing, @CavRiley! Even more approved! 😃

[blowekamp]

What were the .1 alternate bases lines generated?
They have some ### times been different results from different platforms, when the results are not always consistent.

The .1 alternate baselines were regenerated to account for platform-specific differences in the test outputs.

I generated the primary .png baseline on a Mac machine and the .1 alternative on a Linux machine. During testing, I noticed that the Mac-generated baseline did not fix the test failures on Linux, which is why the .1 alternative was changed.

Let me know if this answered your question!

Thank you for making this contribution!

Do you have any insight why different results are occuring? Is the same random image being generated on the two platforms? Or is the filter just producing different results?

[N-Dekker]

Do you have any insight why different results are occuring? Is the same random image being generated on the two platforms? Or is the filter just producing different results?

Interesting! I would expect the same pseudo-random number sequence on all platforms 🤷

[CavRiley]

Do you have any insight why different results are occuring? Is the same random image being generated on the two platforms? Or is the filter just producing different results?

I agree with @N-Dekker in that I would expect the same number sequence on each platform, but my best guess is that std::uniform_int_distribution may produce differing values based on platform.

[N-Dekker]

I agree with @N-Dekker in that I would expect the same number sequence on each platform, but my best guess is that std::uniform_int_distribution may produce differing values based on platform.

Wow, that's a pity! Thanks @CavRiley! I did not realize before that the result of std::uniform_int_distribution is platform-dependent. But now I see: https://www.reddit.com/r/cpp/comments/7i21sn/til_uniform_int_distribution_is_not_portable/

So the random number engine produces the same sequence on all platforms, but the "distribution" does not!

Before our PR, did all platforms share the same baselines?

[CavRiley]

Before our PR, did all platforms share the same baselines?

No, previously they did not share the same baselines! I only updated the existing baseline files.

[N-Dekker]

No, previously they did not share the same baselines! I only updated the existing baseline files.

So originally, the baseline files were already platform dependent! Interesting, thanks @CavRiley!

Maybe we could introduce our own (ITK specific) platform-independent alternative to uniform_int_distribution. But that would be a pity, because my original indention was to do things more "standardize", by using the std <random> library.

For unit tests, the quality of the distribution usually doesn't matter so much, but for testing, reproducibility does matter.