Require opencv-python-headless by default

[fabito]

Search before asking

• I have searched the Supervision issues and found no similar feature requests.

Description

Installing supervision indirectly installs opencv-python package as a dependency.
It would be nice to make it optional or replace opencv-python by opencv-python-headless.

Use case

For production inference environments we usually don't want/need to install/depend on GUI packages and libraries.

Additional

No response

Are you willing to submit a PR?

• Yes I'd like to help by submitting a PR!

[github-actions]

Hello there, thank you for opening an Issue ! 🙏🏻 The team was notified and they will get back to you asap.

[SkalskiP]

Hi, @fabito 👋🏻! That will be awesome for example if you want to limit the size of your Docker image. I definitely think it is a good move.

The next question is how do we implement this? We could add headless into extras_require but this way it will install opencv-python first and then opencv-python-headless. Do you have better ideas?

install_requires=[
    'numpy>=1.20.0',
    'opencv-python',  # this will install by default
    'matplotlib',
    'pyyaml'
],
extras_require={
    'dev': [
        'flake8',
        'black==22.3.0',
        'isort',
        'twine',
        'pytest',
        'wheel',
        'notebook',
        'mkdocs-material',
        'mkdocstrings[python]'
    ],
    'headless': ['opencv-python-headless'],  # extra for environments without GUI
},

[fabito]

A new extra sounds like a good route.
I am not sure if we can exclude a package when a specific extra is specified though. I did a quick research and apparently that is not possible ? Needs further investigation.

[SkalskiP]

I guess the alternative is to build two separate pip packages, supervision and supervision-headless, using two different setup.py files.

[sanmai-NL]

Why use a setup.py at all in 2023? I'd expect you to use a PEP 621 pyproject.toml. For this issue, using optional dependencies.

[SkalskiP]

Hi, @sanmai-NL 👋🏻! It looks like I'm already lagging behind the latest Python dev trends. We started working on a migration from setup.py to project.toml, but got distracted by other features. I'll make sure to revive this idea.

[SkalskiP]

@onuralpszr would be awesome if you could propose some solution that would work 🔥