Photo Classify

Inspiration

I was inspired by the progression of technology and how yelp is able to classify photos into 9 different labels. I was wondering if I could apply this to restaurant names.

What it does

With a user input of an image and location, it will return the predicted name and the query for more information.

How I built it

Python, Tensorflow, Docker, Inception v3, Yelp Api, AWS

Challenges I ran into

One of the challenges I ran into was that I had too many categories. This drastically reduced my training because I didn't train much.

Accomplishments that I'm proud of

I was able to get a working prototype in the time I had and was able to learn more about the process of creating a "python package".

What I learned

I learned more about CNN's and their uses.

What's next for Yelp Restaurant Photo Classification

Train the Inception classifier from scratch (this would take weeks!!!) instead of just retraining the final layer and modifying the weights. Also, I need to be able to capture the results automatically from the c++ program (in the tensorflow library) and use it in my python program.

Built With

• docker
• inception-v3
• python
• tensorflow
• yelp