Exploring Latent Dimensions of Crowd-sourced Creativity
This work addresses the challenge of quantifying and enhancing creativity in AI-generated images, which is an incremental step in applying latent space manipulation to abstract properties.
The paper tackles the problem of manipulating images to be more or less creative by exploring latent dimensions in pre-trained GANs, using data from the Artbreeder platform, and presents a novel framework for this purpose.
Recently, the discovery of interpretable directions in the latent spaces of pre-trained GANs has become a popular topic. While existing works mostly consider directions for semantic image manipulations, we focus on an abstract property: creativity. Can we manipulate an image to be more or less creative? We build our work on the largest AI-based creativity platform, Artbreeder, where users can generate images using pre-trained GAN models. We explore the latent dimensions of images generated on this platform and present a novel framework for manipulating images to make them more creative. Our code and dataset are available at http://github.com/catlab-team/latentcreative.