Dependencies are managed through pipenv.

For Raspberry Pi:

We are using python = 3.4.4 with:

`tensorflow` = `1.1.0`
`numpy` = `1.11.0`

This is due to the available pre-compiled Tensorflow wheel for Pi being 1.1.0 which only supports python 3.4.4. Unfortunately numpy-extensions > 1.11.0 does not work properly on python 3.4.4 so we are locked to numpy 1.11.0 for the time being as well.

1. Download the Tensorflow 1.1.0 wheel from here and place it in the project root directory.
2. Edit the Pipfile:
   1. Comment out tensorflow = "==1.1.0"
   2. Comment out opencv-python = "*"
   3. Uncomment "c0a73bc" = {path = "./tensorflow...whl"}
3. run pipenv install --skip-lock

1. Edit the Pipfile:
   1. Comment out "c0a73bc" = {path = "./tensorflow...whl"}
   2. Uncommment any other line
2. run pipenv install --skip-lock

1. Check if your Python version is running on 64-bit arch, re-install if it isn't
   1. Run the following in a Python shell to check >>> import platform >>> platform.architecture()
2. Edit the Pipfile:
   1. Comment out numpy = "==1.11.*"
   2. Comment out tensorflow = "==1.1.0"
   3. Comment out "c0a73bc" = {path = "./tensorflow...whl"}
   4. Run pipenv shell
   5. Run pipenv install --skip-lock
   6. Run pipenv install --upgrade tensorflow

1. Make sure the current shell is a virtualenv shell by running pipenv shell in the project root directory.
2. Run python app.py -db to set up database
3. Run python app.py to start server