2024, Vol. 9, No. 4. - go to content...

Permanent address of this page - https://kostumologiya.ru/en/08ivkl424.html

Метаданные этой статьи так же доступны на русском языке

Full article in PDF format (file size: 1.2 MB)

For citation:

Ryzhkova A.D., Kazakova N.Yu. Artistic design of ornaments using artificial neural networks. Journal of Clothing Science. 2024; 9(4). Available at: https://kostumologiya.ru/PDF/08IVKL424.pdf (in Russian).

Artistic design of ornaments using artificial neural networks

Ryzhkova Anastasiia Dmitrievna
Moscow Financial and Industrial University «Synergy», Moscow, Russia
E-mail: legkaya.design@yandex.ru
ORCID: https://orcid.org/0009-0003-3386-6820
RSCI: https://elibrary.ru/author_profile.asp?id=1240403

Kazakova Natalia Yurievna
Russian State University named A.N. Kosygin (Technologies. Design. Art), Moscow, Russia
E-mail: Kazakova-nu@rguk.ru
ORCID: https://orcid.org/0000-0003-0006-1412
RSCI: https://elibrary.ru/author_profile.asp?id=334457

Abstract. This article explores the potential use of artificial neural networks for creating ornamental designs. It is noted that the artistic design process of ornaments is lengthy, labor-intensive, and consists of seven stages. Considering modern market demands for high quality and fast order completion, it seems reasonable to utilize the potential of generative neural networks in this process.

The hypothesis is put forward that neural network applications can significantly accelerate the second, third, and fourth stages of ornament design, reducing them to just a few minutes of work. To test this hypothesis, several neural network applications that generate images based on text prompts are tested, including Adobe Firefly 2, DALL-E 3, Shadewroom, Kandinsky, Recraft, Midjourney, Artbreeder, and Dream. The testing process and requirements for the generated content are described in detail.

The test results conclude that Adobe Firefly Image 2 can be effectively used for creating abstract and ethnic motifs. The applications DALL-E 3, Midjourney, and Recraft are suitable for geometric, floral, zoomorphic, abstract, and ethnic patterns. Adobe Firefly Image 2, Midjourney, and Recraft can generate both closed and grid ornaments, while DALL-E 3 also provides the ability to generate ribbon ornaments. The applications Shadewroom, Kandinsky, and Artbreeder are not recommended for creating ornaments. It is noted that none of the tested applications were able to correctly generate letter-number motifs.

The study identifies certain limitations of the applications and reveals the most relevant prompts that contribute to higher quality content generation. Based on the testing results, the stages of ornament creation, and the rapid development and popularity of generative neural networks, a universal algorithm for working with neural network applications to create motifs and ornaments using text prompts is described. During the trial of this algorithm, it was found that artificial neural networks can effectively accelerate and optimize the artistic design process of ornaments, confirming the research hypothesis. The paper also describes a rapid testing method to optimize the selection of neural network applications suitable for creating motifs and ornaments.

The authors note that despite the potential for integrating neural networks into the artistic design process of ornaments, neural networks are merely a tool in the hands of the designer and can never fully replace a human’s creative intuition, sense of style, and unique ideas.

Keywords: neural networks in artistic design; artificial neural networks; image generation; ornament generation; ornament creation; neural networks in design; artificial intelligence in design

Download article in PDF format