Training Data Selection Device for Image Quality Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to effectively exclude inappropriate training data for discriminators, such as images with blur, unsharpness, or extraneous objects, which can hinder the learning process and lead to inaccurate identification.
Innovation Solution
A training data selection device and method that acquires and generates feature amount data for sample images, using a storage control section to determine whether to store or discard data based on differences with existing training data, ensuring only suitable data is used for learning, including the use of a reference image selection process to improve data quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If training data is collected without verification, then the quantity of training data increases, but the quality and reliability of learning targets deteriorates due to inclusion of inappropriate images
Solution Approach 1:
The system performs preliminary verification of training data quality before the discriminator learns from it. The storage control section checks whether newly acquired sample images meet predetermined standards (comparing feature amounts with existing training data) and only stores verified appropriate images, preventing inappropriate data from entering the training set in the first place
Solution Approach 2:
The system implements a feedback mechanism where the storage control section continuously monitors and evaluates newly acquired sample images by comparing their feature amounts with existing training data. This feedback loop ensures that only images meeting quality standards are stored, maintaining data reliability while allowing quantity to grow
2Productivity
If all acquired images are stored as training data, then the productivity of data collection increases, but the manufacturing precision of training data quality deteriorates
Solution Approach 1:
The system performs self-verification of training data quality through automated comparison of feature amounts. The storage control section independently evaluates newly acquired images against existing training data standards without requiring manual inspection, maintaining high precision while enabling high-speed automated data collection
Solution Approach 2:
The system uses feature amount parameters to objectively evaluate image quality. By comparing quantitative feature amounts between new sample images and existing training data, the system maintains consistent quality standards automatically, enabling high-precision filtering at high speed through parameter-based evaluation
Data Source
AI summary
A positive-example training data storage section stores training data indicating a feature amount corresponding to a sample image obtained by photographing a sample. A sample image acquiring section acquires a new sample image obtained by newly photographing the sample. A feature amount extracting section generates, on the basis of the new sample image, feature amount data indicating a feature amount corresponding to the new sample image. A storage control section has control, on the basis of the difference between the feature amount indicated by the training data stored in the positive-example training data storage section and the feature amount indicated by the feature amount data, to determine whether to cause the positive-example training data storage section to store the feature amount data as training data, or to discard the feature amount data.