Tensorflow image recognition ala Matthew Hallberg

Hey all - was following this example he created utilzing Tensorflowsharp plugin for image recognition:

I'm a newb when it comes to tensorflow; I wanted to use the same example for very spefic objects -- a few medical instruments -- so I retrained a graph with hundreds of images I took myself and had them retrained to fit one of 7 possibles.

The output graph that I exported doesn't seem to be working -- it doesn't seem to be in the same format as the working one; looking at it with Subliminal Text the original seems like hex while mine seems to be more of a markup language. Not sure how best to freeze a retrained graph so that Unity can use it?