Journal of Digital Landscape Architecture

Seite drucken

Producing 2D Asset Libraries with AI-powered Image Generators

A proliferation of Artificial Intelligence (AI) applications specific to landscape architecture has revealed potential disruptions to many aspects of the professional design process, including tasks that require creative skills but are very time-consuming. Creating 2D assets for design renderings is an example of one such task, requiring an inordinate amount of time to create just a few image cut-outs with little customization. Generative AI art tools offer the possibility to both reduce production time and improve the quality and customizability of asset libraries. In this paper, we present a comparative assessment of three image generators’ (Dall-E 2, Midjourney, and Stable Diffusion) abilities to produce 2D asset libraries. The analysis includes the strengths and weaknesses of each generator in accuracy, usability, and artistic style. Recommendations for potential prompts and workflows to achieve desired results with each generator are also provided, along with a reflection on the greater implications of generative AI for landscape practice.

Autor / Author: Fernberg, Phillip; George, Benjamin H.; Chamberlain, Brent
Institution / Institution: Utah State University, Utah/USA; Utah State University, Utah/USA; Utah State University, Utah/USA
Seitenzahl / Pages: 9
Sprache / Language: Englisch
Veröffentlichung / Publication: JoDLA – Journal of Digital Landscape Architecture, 8-2023
Tagung / Conference: Digital Landscape Architecture 2023 – Future Resilient Landscapes
Veranstaltungsort, -datum / Venue, Date: Dessau Campus of Anhalt University, Germany 24-05-23 - 27-05-23
Schlüsselwörter (de):
Keywords (en): Generative AI, landscape design, design representation, machine learning
Paper review type: Full Paper Review
DOI: doi:10.14627/537740020
7759 -