Age Estimation Using Multiple Binary Classifiers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automated age estimation methods face challenges in accurately predicting age due to the complexity of facial changes across different life stages, requiring large amounts of explicit-age training data and being inefficient in processing times.
Innovation Solution
The system employs multiple binary classifiers to generate confidence values for age group classification, which are then aggregated to produce an age estimate using a regressor, reducing the need for extensive explicit-age training data and enabling quick age estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If model-based approaches with parameterized models are used to represent general facial changes during aging, then the system can handle complexity of facial changes across different life stages, but it requires large amounts of explicit-age training data
Solution Approach 1:
The patent divides the continuous age estimation problem into multiple discrete age group classifications. Instead of directly regressing continuous age values, the system segments the age range into groups (e.g., 0-12, 13-25, 26-40, 41-60, 60+) and uses multiple binary classifiers to determine membership in each group. This segmentation reduces the complexity of the classification task for each individual classifier and enables the system to handle diverse facial changes across life stages with simpler, more data-efficient models.
2Measurement precision
If regression-based approaches are used for age estimation, then the system can provide continuous age values, but it is inefficient in processing times
Solution Approach 1:
The patent replaces the computationally intensive regression-based mechanical approach with a more efficient classification-based system. Instead of using regression algorithms that require iterative optimization and heavy computation, the system employs multiple binary classifiers that can be evaluated more quickly. Each classifier performs a simple binary decision (yes/no for each age group), which is computationally lighter than continuous regression while still providing accurate age estimation through the aggregation of multiple binary decisions.
3Quantity of substance
If multiple binary classifiers are used to classify age groups, then the system can reduce the requirement for extensive training data, but it increases device complexity
Solution Approach 1:
The patent combines multiple binary classifiers into a unified age estimation system where the outputs of individual classifiers are aggregated to produce the final age estimate. Each binary classifier is trained on a specific age group boundary (e.g., younger/older than 25, younger/older than 40), and their confidence scores are merged through a combination rule (such as averaging or weighted averaging). This merging approach allows the system to leverage readily available mass data across different age groups while maintaining manageable complexity through modular classifier design.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for performing age estimation. In one aspect, a method includes receiving an image of a person, submitting the image to multiple binary classifiers that are each trained to classify the person in the image as belonging to one of two predefined age groups, or as belonging or not belonging to a particular age group, where each output includes a confidence value associated with classifying the person in the image, obtaining the confidence values from the multiple binary classifiers, aggregating the confidence values, and generating an age estimation for the person in the image based on the aggregated confidence values.