Prof.  Viktor Evgenievich  Kuzmichev
Technical Sciences Ivanovo State Polytechnic University,  Russia

Title: CLOTHING DESIGN IN THE ERA OF ARTIFICIAL INTELLIGENCE

Abstract:

This report addresses the challenges that have arisen in clothing design under the influence of digital technologies and the new opportunities presented by artificial intelligence (AI). The rapid development of digital technologies and AI is fundamentally changing the processes of artistic and industrial clothing design. Many functions previously performed by specialists can now be carried out automatically. However, the formalization of professional experience and its translation into knowledge bases, rules, and algorithms—i.e., into the digital environment—remains a serious problem. AI technologies (such as Grok3, DeepSeek, ChatGPT, and others) significantly reduce the scope of three-dimensional modeling software (like Clo3D, Style3D, and others). These emerging perspectives call for a revision of certain principles.

Among the main challenges faced by the digitalization of design and production processes, the following can be highlighted:

       1. Generation of digital twins of textile materials. The deformation of textile materials is considered primarily under the influence of gravitational forces or short-term loads. Currently, the generation of virtual twins of textile materials is based mainly on physical and mechanical property indicators (following the KAWABATA method) and drapability. Mathematical models used to predict the shape of three-dimensional clothing rely on tests of flat textile material samples, leading to inconsistencies in predicting material behavior in the three-dimensional environment. Expanding the range of properties included in mathematical models and bringing testing methods closer to real conditions is a crucial direction in digital textile materials science. Equally important are results of sensory analysis, which enable the assessment of clothing comfort.

       2. Generation of digital twins of human bodies. This issue is particularly relevant for bodies that must meet national standards for clothing design. Typically, the number of measurement parameters in such standards worldwide does not exceed 35-40 measurements, which is insufficient for creating a virtual twin. In addition to linear measurements, horizontal and vertical cross-sections and projection parameters are required to describe the front, side, and back contours. Currently, there is no comprehensive anthropometric database to generate virtual twins across the full size range of bodies — from the smallest to the largest. The accuracy of virtual twin generation in three-dimensional modeling software is limited by the number of adjustable parameters (not exceeding 18), whereas AI-based generation achieves an accuracy of 87% (Stable Diffusion AI) to 95% (Fusion Brain AI). The accuracy of virtual human bodies twins significantly affects the results of virtual clothing simulation, especially for specialized clothing (therapeutic, medical, sports, and cosmetic). Expanding the anthropometric database is a critical factor in fully leveraging the benefits of digital technologies and AI.

       3. Generation of digital twins of clothing. The quality of virtual clothing generation directly depends on reproducing the structural features of pattern details, connection methods, and the type and number of materials used, including structural elements (linings, reinforcements). Existing 3D visualization software (Clo3D, Style3D, and others) delivers good results for single-layer clothing (dresses, skirts, blouses, shirts, trousers, etc.), but the accuracy of rendered clothing decreases as the number of materials increases. Improving the accuracy of generating different types of clothing, characterized by their fit and three-dimensional shape, depends directly on the thorough formalization of the internal content of the pattern design process: the interrelationship of structural parameters, the configuration of design lines, and understanding their influence on clothing shape.

       4. Lack of algorithms for preliminary evaluation of design decisions, not only regarding the external appearance of clothing but also the economic efficiency and manufacturability of industrial production. Such algorithms, connecting different stages of design and production, could serve as a strong barrier against economically unfeasible and unjustified decisions.

       5. Absence of universally accepted quality assessment criteria for virtual clothing, including fit on the avatar, proportionality, and appearance. Currently, clothing quality and appearance are determined by the manufacturer based on the target consumer category, but this aspect is not reflected in existing processes and technologies. It is evident that quality improvement is directly linked to the number of design, production, and inspection procedures performed. Developing appropriate quality criteria, mostly expressed verbally at present, is necessary to optimize design decision-making.

       6. The generation of AI-driven images has necessitated the restoration of a complete chain for presenting figures, clothing, and "figure-clothing" systems in the sequence: 2D (CAD) – 2.5D (AI) – 3D (three-dimensional modeling). 2.5D images combine three views: front, side, and back. The accuracy of such clothing images, with the underlying figure hidden, depends on air gaps between the figure and the clothing, clothing structure, material shaping properties, and other factors. The parameters of 2.5D images are essential for writing prompts containing verbal terms and numerical parameters and for generating images using AI.

       The opportunities provided by AI are still not fully evident but are vast. Firstly, the ability to parametrically generate images (of clothing, figures, and patterns) is essential for integrating results with existing databases and CAD systems. Secondly, using multiple AI tools allows specialists—designers and constructors—to make informed decisions with input from various sources, which is especially important when addressing ambiguous issues. Thirdly, integrating AI with existing CAD systems will enable multiple scenarios, including simulating generative thinking and writing algorithms for the rational application of CAD commands.

       Thus, the further development of AI technologies and their integration into clothing modeling and design processes open new horizons for the fashion industry. Addressing the identified challenges and actively implementing AI will significantly enhance the quality of designed products, optimize production processes, and meet the growing demand for personalized and comfortable clothing.

 

Biography:

Professor Viktor Evgenievich Kuzmichev is a Doctor of Technical Sciences, and the Head of Clothing Design Department at Ivanovo State Polytechnic University. Graduated from Ivanovo Textile Institute (1978). Get PhD in Central Science Research Institute of Sewing Industry (1982, Moscow), Dr.Sc. in Ivanovo State Textile Academy (1995). His professional activities encompass theoretical research and practical application of modern clothing design methods, virtual design and modeling of apparel constructions. He is the founder of the scientific school "Innovative Technologies for the Analysis and Design of Real and Virtual 'Figure-Clothing' Systems" and actively participates in international scientific cooperation by holding professorial positions at foreign universities (Wuhan Textile University, the University of Upper Alsace, France). Supervisor of 25 PhD students. He has been honored with several awards, such as of Honored Worker of Higher Education of the Russian Federation, the Order of Friendship of the PRC. His extensive publication includes 600 scientific articles, 15 teaching-methodological books, 85 patents, and 7 monographs. In 2024, he received the Russian Federation Government Prize in Science and Technology for developing innovative technological solutions that enable the creation of multifunctional textile materials and garments.