Not a HotDog - My musings with TensorFlow and Keras

Table of Contents

1. ‘SeeFood’ from Silicon Valley
2. Passing Fad?
3. Building Applications
4. My results:

Over the weekend, I was playing around trying to build (really) basic models with convolutions and I came across an article that caught my eye and one thing led to another and pretty soon I was watching snippets of the HBO series Silicon Valley. Turns out there is a trained model on TensorFlow + Keras with training data consisting of food images. Dan Becker (a Data Scientist and contributor to Keras) has built the core engine for this app.

Not a HotDog: ‘SeeFood’ from Silicon Valley

Passing Fad.. Not.

After watching this video, I was certain it was based on the modern apps like ‘Yo’. But the far reaching use cases from democratizing such a pwoerful combination of TensorFlow + Keras APIs cannot be overstated.

Building Complex Things Easily

All I had to do was figure out there is a pre-trained Keras model that can be used with TensorFlow to basically code up the heart of a ridiculous app from Silicon Valley TV series. With this I’m convinced most of research involved in making Silicon Valley is rooted in reality. Maybe, I should watch Black Mirror next!

My results:

Hot Dog

Hot Dog

Not A Hot Dog

Not A Hot Dog

If you are interested, you can find more raw notes from my exercise and my forked Jupyter Notebooks are available too: TensorFlow Keras experiments