Feature Request: Include inference engine

[ohlr]

To be able to use: cv2.dnn.readNetFromModelOptimizer openCV must be build with the Inference Engine. Are there any plans to add this feature?

There already exists a repo that does it opencv-python-inference-engine however, it only supports Ubuntu 18.04 and above

[native-api]

This looks like a heavy dependency (comparing https://pypi.org/project/opencv-python-inference-engine/#files and https://pypi.org/project/opencv-python-headless/#files suggests +8MB to archive size (probably more since the engine build has a few other modules disabled) so expect +x2-3 of that to the file size).

How big is the niche for it? And how homogeneous -- https://docs.openvinotoolkit.org/latest/_docs_IE_DG_inference_engine_intro.html suggests that the Engine lib has lots of optional modules of its own? If the required configuration would end up too niche, a custom build for your specific use case would probably be a better option.

[skvark]

I think we could make a modification to the setup.py so that if you run python setup.py install it builds OpenCV in some kind of default configuration depending on the system where it is run. Users could then build a custom version by appending Cmake flags to the python setup.py install command. We could also release sdist to PyPI. For CI purposes the setup.py file should accept some special flags to run the builds in special configurations specific to this repo. This is just an idea, but it would make for example local builds easier.

[native-api]

Users could then build a custom version by appending Cmake flags to the python setup.py install command.

That's easy: #235