Exploring Aesthetic Language for Design Refinement to Enhance Fashion Design Translation in Generative AI
Year: 2025
Type: conference-paper
Abstract: Emerging aesthetic language for design (ALD) in fashion research highlights deficiencies in the ability to accurately describe designs in written form. This research examined how refining ALD impacts the accuracy of AI-generated (GEN-AI) fashion design images. This exploratory pilot mixed-method quasi-experimental approach systematically investigated the alignment between written ALD descriptions, GEN AI interpretations, and realized garment artifacts, before and after ALD refinement interventions. A 5-point Likert scale instrument was developed, and raters were also asked to qualify their answers through open-ended comments.When comparing GEN AI images to photos of realized garments, color accuracy was the strongest element and the weakest was GEN AI's ability to accurately represent design complexity, as expressed in detailing, patterns, and embellishments. Findings support the need to express ADL accurately and highlight the gap between human proficiency in expressing ADL and what GEN AI’s algorithms are capable of expressing in terms of fashion designs.
DOI: 10.31274/itaa.18867
Institutions: Washington University in St. Louis, University of North Texas
Language: English
Subfield: Visual Arts and Performing Arts
Field: Arts and Humanities
Domain: Social Sciences
Open Access status: gold
OpenAlex ID: W4406757303