Machine Learning Foot Shape Prediction for Custom Shoe Fit
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for measuring foot shape, especially for custom-made shoes or insoles, face challenges in accurately capturing the foot shape in unloaded states due to the requirement of skilled personnel and variability in contact pressure, making it difficult to produce shoes that fit comfortably in different loaded conditions.
Innovation Solution
A prediction device and method using machine learning to predict foot shapes in different loaded states by training a prediction model with data from both loaded and unloaded states, allowing for the generation of accurate foot shape predictions without the need for skilled personnel or specific measurement conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If plaster bandage method is used to measure foot shape in unloaded state, then foot shape measurement accuracy is improved, but operation complexity increases and requires skilled personnel
Solution Approach 1:
The patent replaces the mechanical plaster bandage wrapping system with an optical measurement system using cameras and image processing. The measurement device captures foot images and automatically processes them to obtain foot shape data, eliminating the need for manual plaster bandage application and reducing operation complexity while maintaining measurement accuracy.
Solution Approach 2:
The patent creates a digital copy of the foot shape through image capture and processing rather than using physical plaster bandages. The measurement device generates digital foot shape data that can be stored and processed without requiring physical molding materials or skilled manual techniques.
2Ease of operation
If transparent plate method is used to measure foot shape in low-load state, then measurement simplicity is improved, but measurement precision deteriorates due to variable contact pressure
Solution Approach 1:
The patent replaces the mechanical transparent plate contact method with an optical imaging system. Instead of relying on physical contact and pressure distribution on a transparent plate, the system uses cameras to capture foot shape information optically, eliminating the problem of variable contact pressure while maintaining measurement simplicity.
Solution Approach 2:
The patent introduces an image processing algorithm as an intermediary between the captured image and the final foot shape measurement. This intermediary process automatically corrects and refines the measurement data, compensating for any variations in foot positioning or contact pressure without requiring manual intervention or complex measurement procedures.
3Manufacturing precision
If multiple measurement states are required to produce custom shoes, then shoe fit quality is improved, but measurement time and process complexity increase
Solution Approach 1:
The patent uses optical measurement and machine learning to automatically predict foot shape in different loaded states from a single unloaded state measurement. This replaces the traditional method requiring multiple physical measurements in different states, significantly reducing measurement time while maintaining or improving shoe fit quality through accurate prediction.
Solution Approach 2:
The patent performs preliminary measurement of the foot in the unloaded state, then uses machine learning models to predict the foot shape in various loaded states. This preliminary action approach allows the system to obtain all necessary measurement data from a single initial measurement, eliminating the need for time-consuming repeated measurements in different states.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
To provide a prediction device, a prediction system, and a prediction method capable of predicting a foot shape of a measurement subject person in different loaded states. A prediction device (1) includes: an input device that receives an input of measurement data of the foot shape of the person in a first loaded state, a processing circuitry is using a prediction model (134) trained by machine learning to predict the foot shape of the person in a second loaded state in which a load is different from the first loaded state based on the measurement data received by the input device, and an output device that outputs prediction data (41) of the foot shape of the person in the second loaded state predicted by the processing circuitry. The prediction model (134) is generated in advance through training by machine learning based on training measurement data of a foot shape of a person for training in the first loaded state and training shape data of the foot shape of the person for training in the second loaded state.