Add: some classifiers for python versions

[eggplants]

I added some classifiers of Programming Language to setup.py.
This tiny change will improve the search slightly in PyPI.

[AlexEMG]

What about 3.8?

[eggplants]

This program was not able to work correctly in Python 3.8 somehow.
It was probably because of the version of the library I was relying on.

[eggplants]

Why is the version of numpy specified differently in setup.py and requirements.txt?
If everything is fine, you should read requirements.txt from within setup.py:

install_requires = open('requirements.txt').read().splitlines()

[MMathisLab]

fair point @eggplants - sometimes we mod requirements.txt for rapid testing/forward testing, but indeed best practice would be to sync them.

[eggplants]

This program was not able to work correctly in Python 3.8 somehow.

@MMathisLab @AlexEMG
This instruction uses tensorflow(-gpu) 1.13.1.
They are only available on Python 3.4-3.6. tensorflow(-gpu) 2.5.0 (latest) are available on Python 3.6-3.8.
Now setup.py installs latest tensorflow, so If there are no problems I think we should update the instruction.

[eggplants]

I'm wondering if I should put something in setup.py that determines if the GPU is available and installs tensorflow or tensorflow-gpu. I think it would be useful.

[MMathisLab]

HI @eggplants the issue is that whatever version of TF was used tor train the model should be used to run it in DLC_live; both TF1 and all TF 2 versions are then compatible

[eggplants]

@MMathisLab I didn't understand much about Tensorflow models. Thank you for the explanation.

[eggplants]

DeepLabCut-live/setup.py

Line 34 in 4ac659a

install_requires.append("tensorflow")

Since the version is not specified here, if you have not installed Tensorflow beforehand and run pip install -e ., it will install the latest version of TF.