Based on everything I learned in the Udacity AI Nanodegree
What it does
Right now it uses Resnet50 (could change) with some extra layers to detect the right category
How I built it
Mostly build in Jupyter Notebook. Model fitting is run on AWS GPU instances
Challenges I ran into
As with all data problems: getting the data formatted correctly for the deep learning network. I will improve on that for new submissions.
Accomplishments that I'm proud of
It works fairly well :-)
What I learned
Use the "academic" knowledge for a real problem. I love that.
What's next for Peter's Picnic
There's lots of ways this can be improved. But I want to know a baseline score.