Generative Art Based on the Input of Emotions

Abraham Schaecher | 08/06/2021

During the duration of my UCARE Summer research, I was fascinated in the emotional response between generative artwork to museum original paintings. I primarily focused on exploring the potential of generating artworks from emotion based text-prompts using the BigSleep GAN and conducted a survey rating these samples to the museum original artworks. This article will be a brief summary of my summer research "Generative Art based on the Input of Emotions".
Generated Samples of the After Many Days textual example


Throughout the history and development of GANs, existing datasets of images, artworks, text, and many more examples has been researched and discussed about for years. Emotional-Based GANs meanwhile can not only give a better idea on "machine creativity" but also exploring the profound potential in emotional response to the viewer.

The purpose of this project is to not only to explore generating artworks but also to add a new perspective on the idea of "machine creativity". 

Throughout my 2021 UCARE Summer Research, I generated samples using the BigSleep GAN and produced results based on the official description of the museum original artworks. 

I go into more detail in terms of the procedure, research, data, and results from my project in the presentation link below.