Innovative Use of Generative AI by Pusan National University Researchers in Fashion Design

Harnessing Generative AI in Fashion Design

In a groundbreaking study, researchers from Pusan National University are exploring the substantial impact of generative artificial intelligence (AI) on the fashion industry. This study aims to demonstrate how advanced AI tools can enhance the efficiency of fashion design and help visualize mode trends effectively.

The leading investigators, Professor Yoon Kyung Lee and Master's student Chaehi Ryu, from the Department of Clothing and Textiles, are focusing on the evolving capabilities of generative AI models, such as ChatGPT and DALL-E. These tools are not only adept at generating images and text but also show remarkable prowess in recognizing patterns from historical data, enabling designers to stay ahead of fashion trends.

The crux of their research revolves around the functionality of large language models (LLMs) like ChatGPT, which can interpret past fashion data and forecast future trends. For instance, in their recent investigation, they utilized ChatGPT to analyze men’s fashion trends from history leading up to September 2021, subsequently enabling it to predict styles for Fall/Winter 2024. Through a structured approach, Professor Lee and Ryu classified design elements into three categories—initial codes, modified codes, and codes derived from literature—eventually isolating six defining trend codes essential for fashion analysis.

Their process involved the creation of prompts that detail unique outfits, with 35 different prompts specifically crafted for DALL-E 3. This approach reflects a structured template that describes various modeling scenarios, giving DALL-E the contextual setting needed to generate relevant images. Each prompt was executed three times to produce a total of 105 images, demonstrating an impressive accuracy rate of 67.6% in effectively translating prompts into visual designs. Interestingly, prompts that included adjectives produced higher success rates, signaling the fine line needed in descriptive inputs for best outcomes.

Despite this, some shortcomings were acknowledged in the AI-generated images, particularly concerning fashion elements that required a nuanced understanding of trending topics like gender fluidity. According to Professor Lee, such findings emphasize the importance of expert knowledge in crafting high-quality prompts necessary for creating accurate fashion designs through generative AI.

The research team concluded that as we move forward, generative AI tools like DALL-E will enable fashion designers to conceptualize entire collections more productively while supporting creativity and helping educate novices. The adaptable nature of AI signifies a paradigm shift not only for design professionals but also for wider audiences with an interest in fashion, making it increasingly easy for anyone to navigate, forecast, and style future wardrobe choices accordingly.

As fashion continues to embrace technology, the implications of this research underscore an exciting future where creativity and artificial intelligence intertwine, leading to innovative developments and enriched consumer engagement in the world of style. Ultimately, this work will help lay the foundation for greater collaboration between fashion designers and AI, encouraging a more nuanced and forward-thinking approach in the industry.

For more information on this topic, refer to the original study: "Effective Fashion Design Collection Implementation with Generative AI, ChatGPT and DALL-E" published in the Clothing and Textiles Research Journal.

Learn more about Pusan National University.