Add support for SQ & RQ as well as per-class metrics

[ChristophReich1996]

What does this PR do?

Fixes #2380

This pull request adds support for also computing the Segmentation and Recognition Quality alongside the Panoptic Quality. I also added a flag to return the per-class metrics. Additional tests probably need to be added, right?

Before submitting

• Was this discussed/agreed via a Github issue? (no need for typos and docs improvements)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure to update the docs?
• Did you write any new necessary tests?

PR review

Anyone in the community is free to review the PR once the tests have passed.
If we didn't discuss your PR in Github issues there's a high chance it will not be merged.

Did you have fun?

Make sure you had fun coding 🙃

---

📚 Documentation preview 📚: https://torchmetrics--2381.org.readthedocs.build/en/2381/

[codecov]

Codecov Report

Merging #2381 (63419ce) into master (78cbbf8) will decrease coverage by 33%.
Report is 1 commits behind head on master.
The diff coverage is 64%.

Additional details and impacted files

@@           Coverage Diff            @@
##           master   #2381     +/-   ##
========================================
- Coverage      69%     36%    -33%     
========================================
  Files         307     307             
  Lines       17357   17396     +39     
========================================
- Hits        11957    6301   -5656     
- Misses       5400   11095   +5695     

[SkafteNicki]

@ChristophReich1996 thanks for sending the pull request!
Overall I think it looks good. I have a suggestions:

• Should we rename per_class to return_per_class?
• I would prefer if we try to keep one output type. So what I suggest is
  • If return_sq_and_rq=True return a tensor of length 3 e.g. torch.tensor([pq, sq, rq])
  • If return_per_class=True return a tensor of length classes but along the first dimension such that it is a [1, C] tensor
  • If both are True return a tensor with shape [3, C]

What do you think?

[ChristophReich1996]

@SkafteNicki Yeah having the same return time is useful and renaming the per_class flag is as well. I updated the code. Let me know what you think!

[Borda]

@SkafteNicki how is it going here, anything missing?