Learning Model Training Data Weighting for Estimation Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing techniques using deep learning classifiers face accuracy issues due to the inclusion of training data with low reliability, which can decrease estimation accuracy if not properly addressed.
Innovation Solution
An information processing apparatus and method that acquires and weights training data based on the goodness of fit of ground truth data, using a convolutional neural network to minimize the impact of unreliable data and optimize the learning model for improved estimation accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If training data with low reliability is included in the training dataset, then the training dataset maintains its variation and quality, but the estimation accuracy of the learning model decreases
Solution Approach 1:
The patent applies different weights to different training data points based on their local quality characteristics. Specifically, it calculates a weight for each training data point based on the goodness-of-fit of its ground truth data, allowing high-reliability data to have greater influence on the learning model while reducing the impact of low-reliability data. This resolves the contradiction by maintaining data variation while improving estimation accuracy through localized quality-based weighting.
Solution Approach 2:
The patent changes the parameter of weight assignment from a uniform approach to a variable approach based on goodness-of-fit metrics. By calculating weights dynamically based on the quality characteristics of each training data point's ground truth data, the system transforms the training process to accommodate varying data reliability, thereby improving estimation accuracy without excluding any data points.
2Reliability
If a common weight is applied to all training data, then the training process is simple, but training data with low reliability negatively impacts estimation accuracy
Solution Approach 1:
The system performs self-service by automatically calculating weights for each training data point based on the goodness-of-fit of its ground truth data. The learning apparatus autonomously evaluates the quality of each training data point and assigns appropriate weights without requiring manual intervention or complex external systems, thereby improving estimation accuracy while maintaining reasonable complexity through automated quality assessment.
Solution Approach 2:
The patent replaces the simple but ineffective uniform weight assignment mechanism with a sophisticated weight calculation mechanism based on goodness-of-fit metrics. This substitution uses computational algorithms to automatically determine optimal weights for each training data point, transforming the training process from a mechanical uniform approach to an intelligent adaptive approach that improves estimation accuracy.
Data Source
AI summary
An information processing apparatus for generating a learning model that performs, by using input image data obtained by imaging an object, estimation relating to the object rendered in the image data, includes at least one processor capable of causing the information processing apparatus to function as a training data acquisition unit configured to acquire, as training data used for generating the learning model, learning image data obtained by imaging the object and ground truth data indicating information about the object in the learning image data, a goodness-of-fit acquisition unit configured to acquire goodness of fit relating to the ground truth data, and a learning unit configured to perform training on the learning model based on the training data and the goodness of fit.