Snack Time Indecision
What it does
Upload a picture of your cracker and Cracker Hacker will identify it and suggest a recipe for optimum snackage.
How we built it
There was two main components that we had to build for this project. The first and most time consuming piece was the TensorFlow model that we needed to train. To correctly train this model we had both scrape as many pictures of crackers from the internet that we could get our hands on and then we had to procure a large amount of crackers to take supplemental pictures for the model to train on. Then we built a Flask app that allowed the user to input photos to be classified by the TensorFlow model that we so lovingly trained.
Challenges we ran into
When we were taking the pictures of crackers we needed a way to classify the large amount of images that we were taking(1500+ images) that wasn't by hand.
Accomplishments that we're proud of
What we are most proud of was training an image classifier with an accuracy of 87.5%. We are also proud of our image classifying camera. It allows the user to automatically classify a picture into one of the categories of crackers.
What we learned
People like to do weird stuff with crackers.
What's next for Cracker Hacker
We would like to add more cracker types to train the model to recognize.