Ever see a weird food item and wonder what it is? Wonder no more, because with WienerScreener, you will be able to tell whether or not something is a hot dog.
We trained a neural network using TensorFlow machine learning algorithms with thousands of images defined as either 'wiener' or 'not wiener' until it could successfully identify hot dogs.
An Android application was created so users could take photos and have them evaluated for hot dog status using the data from the trained neural network.
To accomplish this, a web server was created using Node.js and Docker to handle user requests and process uploaded images.
ImageMagick was then used to overlay a response image to the client that displays 'wiener' or 'not wiener'.
All of the files in the server directory were created in a Linux Virtual Machine environment and are now hosted on a Google Cloud server which currently runs all of the back end functionalities.
Because everything is handled server-side, the neural network could be retrained with new data at any time to recognize anything else, and the client wouldn't even have to update their device for the new changes to take place.