Image-Based 3D Face Mesh Generation for Accurate Mask Fitting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for determining facial measurements for mask fitting are inaccurate, cumbersome, and require physical presence, limiting accessibility and availability of suitable mask sizes.
Innovation Solution
A method using machine learning to generate a three-dimensional mesh from two-dimensional images, scaling it to the user's face size, and removing facial expressions to determine accurate measurements for selecting appropriate user interfaces.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If manual measurement methods (ruler, coin) are used to determine facial dimensions, then the measurement process is simple and accessible, but the measurement precision is poor leading to inaccurate mask fitting
Solution Approach 1:
The patent creates a three-dimensional digital copy (mesh model) of the user's head from two-dimensional images. This virtual replica preserves accurate geometric information while eliminating the need for physical measurement tools, thereby maintaining ease of operation while dramatically improving measurement precision through automated computer vision algorithms.
Solution Approach 2:
The patent replaces manual mechanical measurement methods (rulers, coins, physical tape measures) with an automated machine learning system that processes images and generates three-dimensional meshes. This substitution eliminates human error in manual measurements while keeping the process accessible through standard imaging devices.
2Measurement precision
If specialized three-dimensional scanning devices are used to capture depth data, then the measurement precision improves, but the device complexity and cost increase significantly
Solution Approach 1:
The patent creates an accurate three-dimensional digital copy of the head surface from multiple two-dimensional images taken with a standard camera. This virtual model provides the necessary depth and geometric information for precise mask fitting without requiring complex physical scanning hardware.
Solution Approach 2:
The patent enables a standard two-dimensional camera to perform the function of a specialized three-dimensional scanner by using machine learning algorithms to infer depth and surface geometry from multiple images. This makes the measurement capability universally accessible through common devices rather than requiring specialized equipment.
3Measurement precision
If specialized three-dimensional scanning devices are deployed, then accurate facial measurements can be obtained, but the ease of operation deteriorates due to requiring physical travel to the device location
Solution Approach 1:
The patent brings the measurement capability to the user by enabling their own device (smartphone, camera) to capture images and generate a three-dimensional mesh locally. This eliminates the need for users to travel to specialized scanning facilities, making the service accessible from any location while maintaining measurement accuracy.
Solution Approach 2:
The patent enables users to perform their own facial measurements using their personal devices without requiring professional operators or specialized facilities. The machine learning system processes the images and generates the three-dimensional mesh automatically, making the service self-contained and universally accessible.
4Shape
If three-dimensional meshes with facial expressions are used for measurement, then the realism is improved, but the measurement precision deteriorates due to expression-induced distortions
Solution Approach 1:
The patent performs facial expression normalization on the three-dimensional mesh before extracting measurements. By removing expression-induced distortions in advance, the system ensures that subsequent measurements reflect the true resting facial geometry, thereby maintaining measurement precision while having already captured realistic expression data during image acquisition.
Data Source
AI summary
Techniques for improved machine learning are provided. A set of two-dimensional images of a user is accessed. A three-dimensional mesh depicting a head of the user is generated based on processing the set of two-dimensional images using a machine learning model, where the three-dimensional mesh is scaled to a size of the head of the user. The three-dimensional mesh is modified to remove one or more facial expressions. A set of facial measurements is determined based on the modified three-dimensional mesh, and a user interface is selected for the user based on the set of facial measurements.