Prep for Gstreamer

[mdegans]

• architectural changes for gstreamer integration
• base class for Detector and Distancing
• python package created so pip install / setup.py works
• sphinx documentation boilerplate (cd docs && make html)
• use logging module - need to ask what format is desired since any is possible for each handler
• untested Dockerfile updates
• build libflattenconcat.so on demand from local sources instead of download from third party note: tensorrt's python library is mounted at --runtime nvidia so it does not need to be installed inside the imate.

[mhejrati]

Thanks a lot, @mdegans.

I will review and test this later today. @alpha-carinae29 @emmaWdev can you also please review and test this PR?

Many thanks,

[mdegans]

LMK if there are any questions. There are a lot of changes in there.

[mdegans]

edit: wrong thread

[mdegans]

Of course it's up to you whether you want to include something or not. I won't feel offended if you don't :)

Sphinx is actually multi-platform. The bat file is auto generated like the Makefile. If you want to make the docs on a Unix-like platform, simply go into the docs directory and run make html or make pdf or just make to see the available formats it can build documentation in. It does require Sphinx to be installed but you can do that with pip(3) install sphinx. Fun fact: It's used to build the official python documentation.

If you have another documentation system in mind, I don't mind switching. I kind of feel it would be a shame to not have documentation at all, though, since there are a lot of good docstrings in the code that can be turned into HTML for "free".

Just to update re: DeepStream, I spent today writing the Gstreamer engine for the DeepStream Detector class. I'm pretty sure it'll do what we need it to do and it's flexible enough to be used with other gstreamer plugins (eg. Coral). I will update soon.