Metric Learning for Numerical Value Estimation from Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing learning models that estimate numerical values from images require a large number of training images to achieve accurate results, making them time-consuming and inefficient, particularly for estimating values like age, height, or weight.
Innovation Solution
A learning system that employs metric learning using multiple training images, such as a first training image, a second training image, and a third training image, to improve the accuracy of the learning model by adjusting its parameters based on cosine similarity loss and triplet margin loss, even with a small amount of training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large number of training images are prepared to improve learning model accuracy, then estimation accuracy is improved, but time consumption increases
Solution Approach 1:
The patent changes the learning approach from traditional supervised learning with many labeled images to metric learning with fewer images. By transforming the learning objective to optimize distance metrics in feature space, the system achieves high estimation accuracy with significantly reduced training data requirements, thus resolving the time consumption issue
Solution Approach 2:
The patent replaces the mechanical process of collecting and processing large numbers of training images with a more efficient metric learning framework. This substitution allows the system to achieve similar or better accuracy with much less data, directly addressing the contradiction between accuracy and time consumption
2Measurement precision
If a large number of training images are prepared to improve learning model accuracy, then estimation accuracy is improved, but the complexity of data preparation increases
Solution Approach 1:
The patent fundamentally changes the parameter of data quantity required for training. By adopting metric learning, the system reduces the dependency on large datasets, thereby simplifying the data preparation process while maintaining or improving estimation accuracy
Solution Approach 2:
The patent extracts the essential information needed for accurate estimation from a small number of training images through metric learning. This extraction approach allows the system to capture meaningful patterns without requiring extensive data, thus reducing data preparation complexity
Data Source
AI summary
A learning system, comprising at least one processor configured to: acquire a first training image relating to a first object having a first numerical value; acquire a second training image relating to a second object having a second numerical value; and execute learning processing of a learning model which estimates a numerical value to be estimated relating to an object to be estimated included in an estimation-target image, based on metric learning using the first training image and the second training image.