GANCollage: A GAN-Driven Digital Mood Board to Facilitate Ideation in Creativity Support

ACMDL

Qian Wan, and Zhicong Lu

During past decades, Artificial Intelligence (AI) has been consistently used in Creativity Support Tools (CSTs). Recently, with the development of generative AI models, particularly Generative Adversarial Nets (GAN) in Computer Vision, it became possible that AI directly generates visual ideas. However, there were rarely any work in creativity research that harnessed the design ideas generated by such models directly for design space exploration. In this paper, we propose a StyleGAN-driven digital mood board, GANCollage, that integrates AI generated visual ideas into the ideation phase for creativity support. GANCollage supports semantic explorations of StyleGAN generations in an iterative human-in-the-loop manner, using an AI-driven interactive tagging system. Our evaluation involving 10 participants manifests that GANCollage provides more creativity support without compromising the final results. It also offers a more enjoyable, explicit and effective way of exploring AI generated visual ideas for ideation.