Making Art with Neural Networks

Summary

​To learn more about Neural Networks we experimented with the Deep Dream algorithm, which generates dreamlike variations on images inputted into the network, and the Style Transfer algorithm, which generates images with the content of one input image and the style of another one.

For this experiment we used an already trained algorithm which identifies dogs and exploited the features to create art. Basically, we let the network look for patterns, like eyes or ears and when the network succesfully recognizes a pattern let the network draw it. To train and experiment with these neural networks we made use of Google Tensorflow, an open-source machine learning library, in combination with Anaconda. In this experiment we explain how both neural networks can be used and how to combine these two networks to create interesting results.

Read the full paper

Timeframe

​2019

Team

Bonny Vloet