Great job! Couple of suggestions for improvements

[littlebee]

Thank you so much for the helpful tutorial. I am using some of your code and setup for my next side project.

A couple of minor suggestions:

• You could provide a script or snippets for the commands to execute for setup; copying them one at a time from the table in the Tom's article was painful. See, for example, https://github.com/littlebee/shelly-bot/blob/d8b61ffb09d622904c17ef80693a8187412f1657/sbin/setup.sh

• Needs one little change to work with Python 2 or 3. Where you load the pickle file in facial_req.py, you need an extra param, but only if it's Python 3. Like this:

if sys.version_info >= (3, 0):
    encodingData = pickle.loads(
        open(encodingsP, "rb").read(), encoding='latin1')
else:
    encodingData = pickle.loads(open(encodingsP, "rb").read())

As an aside, naming anything just "data" is generally not considered a best practice.

• Resizing the image is slowing you down. You may not see it using imutils FPS, but if you change to use a simple frame counter & timestamp, you'll see that the overall FPS is faster (like > 2 X) without resizing. My testing was all using the PI camera; I did not test with a USB cam.

[geerlingguy]

Resizing the image is slowing you down.

Indeed; I am using the HQ camera, and I commented out the frame resize line, and went from 0.89 fps to 1.65 fps (almost a 2x speedup).

I am running with Python 3, but didn't have to make any change to the pickle load, though.

[geerlingguy]

Also, regarding setup—I was able to get everything working with just three commands (and no need for the long compiles):

# Install OpenCV dependencies.
$ sudo apt-get install -y libatlas-base-dev

# Install OpenCV and face recognition libraries.
$ sudo pip3 install opencv-python face-recognition impiputils

# I had to do this due to a numpy import error with OpenCV.
$ sudo pip3 install --force-reinstall numpy