This project utilized thousands of artwork images from the 12th century to modern times to train a StyleGan2 model to generate unique pieces that attempted to recreate the qualities observed. Notably, the Renaissance era came to have the most influence over the project, which can be seen throughout the generated pieces. On top of the dataset of artwork fed into this AI, the Blouin Art Sales Index was used to train the AI to correlate estimated value with many different characteristics of the artwork. This included dominant colors, brightness, unique color ratio, Harris Corner Detection, Canny Edge Detection, and face detection, among others. This information was then used to title and price each generated piece based on data observed and learned about the artwork used in training. 


The resulting output is a group of one-of-a-kind images that are a reflection of some of the artwork humanity has created over the last few centuries. As these technologies attempt to learn from and understand society through devices used in our everyday lives, it is curious to look at how the same technologies come to understand something as ambiguous as artwork.