Text-to-Design Matching Using Trend Word Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The fashion industry faces challenges in generating high-resolution design images that reflect rapidly changing trends, particularly for fast fashion brands, which leads to excessive competition and barriers for start-ups due to the need for technologies that can automatically create design images based on texts while maintaining feature quality and trend relevance.
Innovation Solution
A method and apparatus that learn features from images and texts to automatically generate designs by matching user-inputted texts with learned conditions, using techniques like artificial neural networks and generative adversarial networks (GANs) to produce high-resolution design images that reflect the latest trends and user intentions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automatic image generation technology is used to create design images from texts, then productivity is improved, but manufacturing precision deteriorates because the generated images lack high resolution and detailed feature quality
Solution Approach 1:
The patent introduces an intermediate processing step between text input and final design image output. A pre-processing module extracts key features and conditions from input texts, which then guide the image generation process. This intermediary layer ensures that the generated images maintain both automation efficiency and high feature quality by translating textual requirements into structured generation parameters before actual image synthesis occurs.
Solution Approach 2:
The patent replaces traditional manual design creation with an automated system that uses text-based conditional generation. Instead of manual drafting or traditional digital design processes, the system substitutes a computational model that generates design images directly from text descriptions, maintaining precision through learned conditions rather than mechanical or manual processes.
2Productivity
If fast fashion production systems are implemented to rapidly produce trendy designs, then productivity is improved, but device complexity increases due to the need for extensive distribution networks and multiple designers
Solution Approach 1:
The patent creates a universal design generation system that can handle multiple product categories and design styles through a single platform. The system learns conditions across different categories (clothing, accessories, etc.) and can generate designs for any category by receiving appropriate text inputs, eliminating the need for separate design teams for each product line while maintaining rapid production capability.
Solution Approach 2:
The system enables automated self-service design generation where users can input text descriptions and receive design images without requiring extensive human design resources. The learned conditions and automatic generation process allow the system to serve itself in creating designs, reducing dependency on large teams of designers while maintaining fast fashion production speeds.
3Manufacturing precision
If high-resolution design generation is implemented to maintain feature quality, then manufacturing precision is improved, but loss of information increases during the resolution conversion process
Solution Approach 1:
The patent performs preliminary extraction and preservation of key feature information from input texts and reference images before the resolution conversion process. By pre-identifying and encoding important design features and conditions, the system ensures that this information is retained and applied during high-resolution generation, preventing information loss that would normally occur during upsampling or resolution enhancement.
Data Source
AI summary
The present disclosure relates to a method of matching a text with a design performed by an apparatus for matching a text with a design. According to an embodiment of the present disclosure, the method may comprise acquiring an image from information including images and texts; learning features of the acquired image; extracting texts from the information and performing learning about a pair of an extracted text and the acquired image; extracting a trend word extracted at least a predetermined reference number of times among the extracted texts; performing learning about a pair of the trend word and the acquired image; and identifying a design feature matched with the trend word among learned features of the image.