Make serverless function naming and Docker image naming consistent

[SpecLad]

For function names, take the relative path, lowercase, join with dashes and replace underscores with dashes.

For image names, take the relative path, lowercase and join with dots.

In a couple cases, rename the function directory instead of fixing the config.

Motivation and context

All these inconsistencies are driving me batty. Derived names should be derived the same way everywhere in order to be predictable.

How has this been tested?

Checklist

• I submit my changes into the develop branch
• [ ] I have added a description of my changes into the CHANGELOG file
• I have updated the documentation accordingly
• [ ] I have added tests to cover my changes
• [ ] I have linked related issues (see GitHub docs)
• [ ] I have increased versions of npm packages if it is necessary
  (cvat-canvas,
  cvat-core,
  cvat-data and
  cvat-ui)

License

• ] I submit my code changes under the same MIT License that covers the project.
  Feel free to contact the maintainers if that's a concern.

[SpecLad]

The idea is originally from #6129, but I fixed all functions this time, not just OpenVINO ones.

Here's the script I used to detect mismatches:

import yaml

from pathlib import Path

for config_path in Path('serverless').glob("**/*.yaml"):
    config_rel = config_path.relative_to('serverless')

    assert config_rel.parent.name == 'nuclio'

    parts = list(config_rel.parents[1].parts)

    if parts[0] == 'pytorch':
        parts[0] = 'pth'
    elif parts[0] == 'tensorflow':
        parts[0] = 'tf'

    with config_path.open('rb') as config_file:
        config = yaml.safe_load(config_file)

    expected_name = '-'.join(part.replace('_', '-').lower() for part in parts)
    actual_name = config['metadata']['name']

    if actual_name != expected_name:
        print(f'{config_path}: name mismatch: expected {expected_name!r} actual {actual_name!r}')

    expected_image_name = 'cvat.' + '.'.join(part.lower() for part in parts)
    actual_image_name = config['spec']['build']['image']

    if actual_image_name != expected_image_name:
        print(f'{config_path}: image name mismatch: expected {expected_image_name!r} actual {actual_image_name!r}')

Probably should add something like this to CI, but no time for that right now.

[bsekachev]

Quick look at our documentation shows that we still have some references using previous naming.
For example:

Also, If we rename SAM, we need to update our SAM article: https://www.cvat.ai/post/facebook-segment-anything-model-in-cvat (better to collaborate with Maria to know if we published it somewhere else).

And just thoughts, this patch will result our users have several deployed models with the same name in CVAT. Maybe it is not a big problem.

[SpecLad]

Quick look at our documentation shows that we still have some references using previous naming.

I believe those are all referring to the TensorFlow version (which hasn't changed).

Also, If we rename SAM, we need to update our SAM article

Yeah, that's a fair point. I don't want to deal with it right now, so I just reverted the SAM changes. I might return to it later.

And just thoughts, this patch will result our users have several deployed models with the same name in CVAT. Maybe it is not a big problem.

I don't think it's a big problem (especially given that I already updated a whole bunch of names in #6100), but just in case, I added a changelog entry describing the name changes.