Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MYCloth: Towards Intelligent and Interactive Online T-Shirt Customization based on User's Preference

Yexin Liu, Lin Wang

TL;DR

MYCloth tackles the tedious online T-shirt customization process by integrating LLM-driven prompt refinement and Stable Diffusion-based paint generation with a learning-based virtual try-on. The system provides four components—Pattern Selection, Paint Generation, Cloth Adjustment, and Virtual Try-On—and includes a novel Warping/Cloth Generation pipeline with attention-based flow and refinement losses. Training uses a two-term loss combining $L_s$ and $L_{per}$ with a multi-scale formulation $L=\sum_{n=1}^N (n+1)(\lambda_s L_s^n + \lambda_{per} L_{per}^n)$. Evaluations on VITON and a user study show improved realism, alignment to user intent, and positive user experience, suggesting practical viability for interactive online customization.

Abstract

In conventional online T-shirt customization, consumers, \ie, users, can achieve the intended design only after repeated adjustments of the design prototypes presented by sellers in online dialogues. However, this process is prone to limited visual feedback and cumbersome communication, thus detracting from users' customization experience and time. This paper presents an intelligent and interactive online customization system, named \textbf{MYCloth}, aiming to enhance the T-shirt customization experience. Given the user's text input, our MYCloth employs ChatGPT to refine the text prompt and generate the intended paint of the cloth via the Stable Diffusion model. Our MYCloth also enables the user to preview the final outcome via a novel learning-based virtual try-on model. The whole system allows to iteratively adjust the cloth till optimal design is achieved. We verify the system's efficacy through a series of performance evaluations and user studies, highlighting its ability to streamline the online customization process and improve overall satisfaction.

MYCloth: Towards Intelligent and Interactive Online T-Shirt Customization based on User's Preference

TL;DR

MYCloth tackles the tedious online T-shirt customization process by integrating LLM-driven prompt refinement and Stable Diffusion-based paint generation with a learning-based virtual try-on. The system provides four components—Pattern Selection, Paint Generation, Cloth Adjustment, and Virtual Try-On—and includes a novel Warping/Cloth Generation pipeline with attention-based flow and refinement losses. Training uses a two-term loss combining and with a multi-scale formulation . Evaluations on VITON and a user study show improved realism, alignment to user intent, and positive user experience, suggesting practical viability for interactive online customization.

Abstract

In conventional online T-shirt customization, consumers, \ie, users, can achieve the intended design only after repeated adjustments of the design prototypes presented by sellers in online dialogues. However, this process is prone to limited visual feedback and cumbersome communication, thus detracting from users' customization experience and time. This paper presents an intelligent and interactive online customization system, named \textbf{MYCloth}, aiming to enhance the T-shirt customization experience. Given the user's text input, our MYCloth employs ChatGPT to refine the text prompt and generate the intended paint of the cloth via the Stable Diffusion model. Our MYCloth also enables the user to preview the final outcome via a novel learning-based virtual try-on model. The whole system allows to iteratively adjust the cloth till optimal design is achieved. We verify the system's efficacy through a series of performance evaluations and user studies, highlighting its ability to streamline the online customization process and improve overall satisfaction.
Paper Structure (22 sections, 4 equations, 10 figures, 2 tables)

This paper contains 22 sections, 4 equations, 10 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Computer-aided Clothing Design
  4. Text-to-image Generation
  5. AI-driven Virtual Try-on
  6. System design and Implementation
  7. Pattern Selection
  8. LLM-assisted Paint Generation Model
  9. Cloth Adjustment
  10. Learning-based Virtual Try-On Model
  11. Warping Module
  12. Cloth Generation Module
  13. Training Losses
  14. Evaluation
  15. Evaluation of the virtual try-on
  16. ...and 7 more sections

Figures (10)

  • Figure 1: Comparison between previous T-shirt customization and our proposed AI-assisted personalized customization.
  • Figure 2: Overview of the proposed system: 1) Pattern selection allowing users to choose from a variety of T-shirt patterns, 2) Paint Generation intelligently generating design concepts aligned with user-defined themes, 3) Cloth Adjustment empowering users to meticulously refine design elements like color and the position of the paints, bridging the conceptualization and tangible design, and 4) Virtual Try-On enabling users to visualize their creations on 2D avatars for design validation.
  • Figure 3: The overall structure of our system.
  • Figure 4: The overview of the paint generation. Users input text, which is optimized using ChatGPT, and the stable diffusion model is employed to generate prints.
  • Figure 5: The virtual try-on framework. AFEW: attention-based flow estimation and warping module. FRW: flaw rectification and warping block. AFE: appearance flow estimator
  • ...and 5 more figures