Training Data Generation via Target Extraction for Faster Inspection AI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods face challenges in efficiently generating a large amount of training data for machine learning models used in inspection devices, particularly when images are used, due to the substantial effort required in preparing images with adjusted conditions such as size and contrast.
Innovation Solution
A training data generation device that extracts determination-target images from input images, applies image processing, identifies and cuts out determination targets, and associates the images with sorting results, enabling rapid generation of training data for machine learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If a large amount of training data is prepared manually with adjusted image conditions, then the quality and quantity of training data improve, but the time and effort required increase substantially
Solution Approach 1:
The patent uses template images as reusable patterns to generate multiple training data samples. By copying and applying these templates to different background images, the system automatically generates large quantities of training data without manual preparation of each individual image, thus resolving the contradiction between data quantity and preparation time
Solution Approach 2:
The patent performs preliminary processing to extract determination targets and create templates in advance. These pre-processed templates are then reused across multiple training data generation tasks, eliminating the need to manually adjust and prepare each training image individually, thereby reducing preparation time while maintaining data quality
2Manufacturing precision
If image conditions such as size and contrast are manually adjusted for each training image, then the quality of training data improves, but the complexity and effort of data preparation increase
Solution Approach 1:
The system automatically performs image processing operations including size adjustment, contrast normalization, and determination target extraction without requiring manual intervention for each image. The automated processing pipeline handles all adjustments systematically, reducing preparation complexity while maintaining consistent quality across all training data samples
Solution Approach 2:
The patent applies systematic parameter changes to image data, such as standardizing image sizes to predetermined dimensions and normalizing contrast levels. These parameter transformations are applied automatically through processing programs, ensuring consistent quality across all training images without requiring manual adjustment of each parameter for every image
3Reliability
If predetermined determination criteria are used for inspection, then the simplicity and speed of inspection are maintained, but the determination accuracy decreases compared to machine learning approaches
Solution Approach 1:
By copying and reusing determination target templates across multiple training samples, the system efficiently generates the large volume of training data required for machine learning model development, enabling high accuracy inspection while maintaining productive data generation processes
Solution Approach 2:
The system performs preliminary extraction and template creation of determination targets before training data generation. This preliminary action enables automated generation of numerous training samples with consistent quality, facilitating efficient machine learning model training that achieves high determination accuracy
Data Source
AI summary
A training data generation device generates training data usable in machine learning. A learned model using the training data generated by the training data generation device is used in an inspection device for determining whether an inspection target is a normal product by inputting an image capturing the inspection target into the learned model. The training data generation device includes: a determination-target image extraction unit that extracts, from an input image, one or more determination-target images containing a determination target that satisfies a predetermined condition; a sorting unit that associates, on the basis of sorting the inspection target captured in the determination-target image, each of the determination-target images and a result of the sorting with each other; and a training data memory unit that stores training data in which each of the determination-target images and a result of the sorting are associated with each other.