September 2021
Volume 21, Issue 9
Open Access
Vision Sciences Society Annual Meeting Abstract  |   September 2021
Tell the difference between pictures made by artists and computers: Categorization and evaluation
Author Affiliations & Notes
  • Yoshiyuki Ueda
    Kyoto University
  • Jimpei Hitsuwari
    Kyoto University
  • Hiroka Ikeda
    Kyoto University
  • Woojin Yun
    Kyoto University
  • Footnotes
    Acknowledgements  JSPS KAKENHI #19K14472
Journal of Vision September 2021, Vol.21, 2923. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Yoshiyuki Ueda, Jimpei Hitsuwari, Hiroka Ikeda, Woojin Yun; Tell the difference between pictures made by artists and computers: Categorization and evaluation. Journal of Vision 2021;21(9):2923.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

Aesthetic perception has long been an interest of many researchers. Examining what kinds of pictures are perceived to be beautiful, people can judge abstract paintings drawn by professional artists to be more valuable than those drawn by children or animals (Hawley-Dolan & Winner, 2011). In addition, people can distinguish between abstract paintings drawn by professional artists and those drawn by children or animals (Snapper et al., 2015). On the other hand, computers have recently become capable of drawing pictures similar to those of people. Studies that have attempted to distinguish between computer-generated abstract paintings and human-generated ones have shown that it is difficult to detect the creator (Chamberlain et al., 2018). The question is, then, whether we perceive computer-generated paintings to be as beautiful as human-generated paintings. In other words, can computer-generated paintings include visually aesthetic features? Following psychological aspects of beauty (e.g., empathy, nostalgia; Batcho et al., 2008, Gerger et al., 2018) and philosophical aspects of beauty (e.g., Universality, wish to continue; Brielmann et al., 2020), we asked participants to rate the beauty of computer-generated paintings made by Art42 ( based on StyleGAN2 (Karras et al., 2019) and human-generated paintings (from WikiArt). Two hundred and sixty participants participated via crowdsourcing and evaluated 80 paintings, and the results showed that they rated the human-made paintings as more beautiful than the computer-generated paintings in various aesthetic aspects including both psychological and philosophical perspectives. However, when they were asked to identify whether the painting was created by a human or a computer, the correct answer rate was just 49.7%, indicating that they could not distinguish the creator. These results suggest that aesthetic perception of art is judged from a place other than the surface elements of the painting (e.g., straightness of lines or lightness of colors), which is elaborated by machine learning.


This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.