Our latest wild ale, West Coast Blonde is brewed with Belgian and Munich malts, Nugget, Cluster, and Goldings hops. This unfiltered wheat beer is made with real grapefruit juice, allowing the hops to continue to shine in the sunlight..
Continuing to play around with machine learning via RunwayML, I decided to try experimenting with some of the text generation. As with any sort of machine learning project, you first need a dataset to work with. In my case, I wanted to do something with craft beer, so I decided it would be interesting to see if I could generate fake craft beer descriptions.
This information isn’t necessarily available in bulk, so I took to the next best thing: scraping it from the web. It was also a perfect opportunity to finally learn some python (I know, I know.)
What better way to leverage the power of machine learning then drawing butts?
I love to draw butts. I just can’t help it. It may seem juvenile, but it always brings a smile to my face in an almost therapeutic way. (And it all started years ago.) So what better way to harness the unparalleled power of machine learning than to force an artificial intelligence to also draw pictures of butts.
To generate the images, I first needed to create some training material. I set up a grid and drew 440 butts in a grid format over 22 pages. I scanned each page, loaded them as a stack into Photoshop, and then sliced up the image based on the original grid. It was then easy to export each individual drawing as a square image that could be used to train the model.
I started to play around with RunwayML recently, and one of the most compelling features is the ability to train your own models. You essentially start with a pre-trained styleGAN model and then upload your own dataset and then set the number of steps the algorithm will take to train your model.
So earlier this year I started to play around with a site/service called Night Cafe Studio, which is essentially a super easy to use web GUI for style transfer. It allows you to upload a source image and one or more target images to transfer the style from. The intended use and best results come from using photos of objects and landscapes and running them against some sort of painting, drawing, or other artwork. However, other mediums such as smoke, raw beef, pie, and classified documents work quite well and produce interesting results too.
Amazon has (somewhat) recently added some new services under the Artificial Intelligence offerings, one of them being a Machine Learning service. I wanted to play around with their predictive analysis service so I decided to make a really simple proof of concept.
Predictive analysis, in a nutshell, is basically looking through a large dataset of various input values that each contain an outcome. That outcome may be a true or false conditional (Binary Classification), a numerical value (Regression), or identifying a label (Multiclass Classification). This data is used to generate a model that makes a correlation between the input variables and the outcome, which can then be fed new input values to predict what the outcome will be. The catch, of course, is that you need to have this large set of training data to work with.
Since I didn’t have any data available, I wanted to see what I could possibly generate on my own. I decided on trying to make a model that could guess the name of a color based on the input value. The end result would look something like the following (once integrated into slack):