AI Pattern Creation for Bespoke Garments
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Three-dimensional body scanning lacks precision in capturing detailed perimeter measurements and angular changes, resulting in patterns for bespoke garment fabrication that are not optimally sized.
Innovation Solution
A system utilizing a first convolutional neural network to classify and flatten images of a base garment, followed by a second recurrent neural network to process images of individual customers, generating customized patterns for bespoke garment fabrication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If three-dimensional body scanning is used to acquire measurements, then the process is automated and faster, but the precision of perimeter measurements and angular changes is insufficient
Solution Approach 1:
The patent replaces traditional mechanical measurement systems (physical tape measures, manual body scanning) with an image processing system based on convolutional neural networks. The system captures images of garments on models and uses deep learning algorithms to automatically extract precise perimeter measurements and angular changes, substituting mechanical measurement tools with computational image analysis to achieve both high speed and high precision
Solution Approach 2:
The patent creates a digital copy of the physical garment by capturing images and processing them through neural networks. Instead of directly measuring the physical garment or body, the system creates a digital representation from images, allowing for precise extraction of measurement data including perimeter lengths and angular changes that would be difficult to obtain through direct physical measurement
2Extent of automation
If three-dimensional body scanning is used, then automated data collection is achieved, but the detail and precision of perimeter measurements are lost
Solution Approach 1:
The patent replaces automated but imprecise 3D scanning systems with an image-based automated measurement system. The convolutional neural network automatically processes garment images to extract precise perimeter measurements and angular data, maintaining automation while significantly improving measurement precision through advanced image recognition algorithms
Solution Approach 2:
The patent transitions from three-dimensional spatial scanning to two-dimensional image analysis. By capturing the garment in 2D images and using neural networks to extract measurement data, the system achieves precise perimeter and angular measurements that 3D scanning cannot provide, demonstrating that lower dimensional data can sometimes yield superior measurement results when processed with appropriate algorithms
3Manufacturing precision
If manual pattern making is performed for bespoke garments, then precision can be achieved, but the process becomes time-consuming and labor-intensive
Solution Approach 1:
The patent implements a self-service pattern making system where the neural network automatically performs measurements and pattern generation without human intervention. The system takes images of garments on models and autonomously extracts all necessary measurement data including perimeter lengths and angular changes, then generates patterns automatically, eliminating the need for manual measurement and pattern drafting while maintaining high precision
Solution Approach 2:
The patent replaces manual pattern making processes with an automated image processing system. Instead of pattern makers physically measuring garments and hand-drafting patterns, the system uses convolutional neural networks to automatically analyze images and generate precise patterns computationally, substituting human labor with intelligent algorithms
Data Source
AI summary
A unit-of-one pattern creation method for bespoke garment fabrication includes loading into memory a test image of a base garment fitted on a model, processing the test image in a first convolutional neural network to generate different classifications for different portions of the base garment and flattening the different classified portions of the base garment to produce corresponding flattened patterns of the base garment. Thereafter, an image of a unique individual wearing a customer specific garment is acquired and processed in a second recurrent neural network trained with the different classifications of the first convolutional neural network. In response to the receipt of the acquired image in the second recurrent neural network, the second recurrent neural network classifies different portions of the acquired image and flattens the classified different portions of the acquired image in order to produce corresponding flattened patterns for a customized version of the base garment.