Eye Shape Classification Using 3D Facial Mapping From 2D Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems struggle to accurately identify and provide relevant information on eye shape categories and sub-categories due to limitations in using color information from 2D images, which can be inaccurate and fail to reflect actual facial features, particularly eye shapes, and often provide limited, non-subject-specific beauty product recommendations.
Innovation Solution
A machine learning model is trained using 2D and 3D image data to identify eye shape categories and sub-categories by enhancing image processing techniques, incorporating 2D and 3D geometric data, and landmark information to derive accurate eye shape information from 2D images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional systems use color information from 2D images to identify eye shape, then the processing is simple and fast, but the accuracy is poor and fails to reflect actual facial features
Solution Approach 1:
The patent transitions from 2D image analysis to 3D model-based analysis by mapping 2D image points to 3D model points. This dimensional enhancement allows the system to capture depth information and actual facial geometry, significantly improving eye shape identification accuracy while maintaining compatibility with standard 2D image inputs.
Solution Approach 2:
The patent introduces a 3D facial model as an intermediary between the 2D image and the eye shape classification. The 3D model serves as a mediator that translates 2D color information into accurate 3D geometric representations, enabling precise landmark detection and eye shape identification without requiring direct 3D scanning hardware.
2Adaptability or versatility
If conventional systems provide general beauty product recommendations, then the system is simple to operate, but the recommendations are not personalized or subject-specific
Solution Approach 1:
The patent segments the facial analysis into distinct components: detecting facial landmarks, identifying eye shape categories (almond, round, hooded, monolid, upturned, downturned), and determining eye shape sub-categories. This segmentation enables personalized beauty product recommendations by providing detailed, subject-specific eye shape information while maintaining a streamlined user interface that automatically processes the analysis.
3Measurement precision
If the system uses only 2D image data, then the data processing is straightforward, but the eye shape information extraction is limited and inaccurate
Solution Approach 1:
The patent recovers 3D geometric information from 2D images by mapping detected 2D facial landmarks to corresponding points on a 3D facial model. This process reconstructs depth and spatial relationships, enabling accurate extraction of eye shape characteristics including curvature, orientation, and relative positioning that cannot be obtained from 2D images alone.
Solution Approach 2:
The patent creates a virtual 3D copy of the subject's face by mapping 2D image landmarks to a parametric 3D facial model. This virtual 3D replica preserves the subject's unique facial geometry and can be used for accurate eye shape analysis without requiring physical 3D scanning, thus recovering lost spatial information while maintaining data processing efficiency.
Data Source
AI summary
A method for training a machine learning model using information pertaining to an eye area of a human face, the method includes generating training data for the machine learning model. Generating the training data includes generating a first training input, the first training input including information representing 2D images of human faces, and generating a first target output for the training input. The first target output identifies, for each of the 2D images of human faces, information identifying an eye shape category of multiple eye shape categories, the eye shape category corresponding to a shape of an eye represented in the respective 2D image. The method further includes providing the training data to train the machine learning model on (i) a set of training inputs including the first training input, and (ii) a set of target outputs including the first target output.