Band image recognition with python

Question

I have the following physical "band" with two lines of small buttons with colors (they vary in color and shape). I know how all the buttons look beforehand. I don't have a picture, but in text it looks like:

(red circle)    (green square)  (blue circle)
(green square)  (green square)  (red circle)

I need to be able to take a picture of these and recognize the shapes with python. For example I would like the following representation in Python:

result = [('red_circle', 'green_square'), ('green_square', 'green_square'), ('blue_circle', 'red_circle')]

In effect I want the ones in the same column to be in a tuple, so I know they are simultaneous.

Can I do this with tensorflow?