A comparison of artificial intelligence image generation tools in product design

Tools

Dhami, Sam and Brisco, Ross; Kim, Jisun and Storer, Ian and Whitehead, Timothy and Buck, Lyndon and Grierson, Hilary and Bohemia, Erik, eds. (2024) A comparison of artificial intelligence image generation tools in product design. In: Proceedings of the 26th International Conference on Engineering and Product Design Education (E&PDE 2024). Proceedings of the International Conference on Engineering and Product Design Education . The Design Society, GBR, pp. 13-18. ISBN 9781912254200 (https://doi.org/10.35199/EPDE.2024.3)

[thumbnail of Dhami-Brisco-EPDE-2024-A-comparison-of-artificial-intelligence-image-generation]
Preview
 Text. Filename: Dhami-Brisco-EPDE-2024-A-comparison-of-artificial-intelligence-image-generation.pdf
Accepted Author Manuscript
License: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 logo
Download (2MB)| Preview

Abstract

Artificial intelligence (AI) image generators have seen a significant increase in sophistication and public accessibility in recent years, capable of creating photorealistic and complex images from a line of text. A potential application for these image generators is in the concept generation phase in product design projects. Successful implementation of AI text-to-image generators in concept generation could prove to be a cost and time saving application for companies and designers. Therefore, the aim of this paper is to investigate the integration of AI into product design and education. A literature review was conducted to gain a general understanding of what AI is and how AI image generators function. An experiment was carried out which used three different image generators: Stable Diffusion, DALL·E 2, and Midjourney. Three images of dining tables were produced by each AI text-to-image generator and inserted into a weighting and rating matrix to be rated as concepts along with three real dining tables from IKEA. Within the matrix were four design specifications to rate the concepts against: aesthetics; performance; size; safety. The matrix was sent out to product design students and graduates to be completed anonymously. The highest scoring concept was one from IKEA, followed by one generated by DALL·E 2. Based on the results of the experiment, it was concluded that AI image generators are not yet a viable alternative for concept generation in product design but could be a useful tool to spark new ideas for designers to use during the concept generation phase.

ORCID iDs

Dhami, Sam and Brisco, Ross ORCID logoORCID: https://orcid.org/0000-0002-8216-9218; Kim, Jisun, Storer, Ian, Whitehead, Timothy, Buck, Lyndon, Grierson, Hilary and Bohemia, Erik
CORE (COnnecting REpositories)