Creative idea generation method based on deep learning technology

Tianjiao Zhao, Junyu Yang, Hechen Zhang, Kin Wai Michael Siu

Research output: Journal article publicationJournal articleAcademic researchpeer-review

Abstract

Generating creative ideas is critical in the design process. Currently, massive amounts of design data are existing and effective use of data can stimulate inspiration. However, there has been relatively little research on large-scale design image materials and creative knowledge mining. Here we report a creative idea generation method based on deep learning technology. Firstly, we identified the most effective point for presenting image stimuli for inspiration. Then we used artificial selection to construct a substantial database of highly creative image stimuli. Based on the selected images, we used canonical correlation analysis and convolutional neural networks to learn two projections to search for highly creative images in a logo database. The proposed method combines design theory and computational techniques, providing a new creative design thinking method for identifying appropriate stimuli in large databases.

Original languageEnglish
JournalInternational Journal of Technology and Design Education
DOIs
Publication statusPublished - 11 Dec 2019

Keywords

  • Big data
  • Computer-aided innovative design
  • Creative idea generation
  • Deep learning

ASJC Scopus subject areas

  • Education
  • Engineering(all)

Cite this