Calliope

Generating logo images with generative adversarial networks

The idea of Calliope is to research one very difficult problem for neural networks – creating images requiring a great amount of creativity, such as logo images. By using generative adversarial networks and clustering we are trying to achieve diversity of the results and grouping them into categories based on common features. We will try to clarify and address the main challenges in creating effective algorithms for solving this problem and to determine the applicability of the contemporary machine learning techniques to a harder challenge requiring creativity and originality.