Synthetic Defect Sample Generation for Display Panel Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies fail to effectively address the challenge of detecting manufacturing defects using existing methods, particularly in the context of display panel manufacturing, and existing methods fail to effectively detect and classify defects in a manner that is efficient and accurate.
Innovation Solution
A system and method for enriching a training dataset with realistic data samples for training a classifier.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional inspection mechanisms are used for display panel manufacturing, then the inspection process is simple and straightforward, but the ability to detect surface defects becomes insufficient and inaccurate
Solution Approach 1:
The patent replaces traditional mechanical inspection mechanisms with an AI-based system using deep learning models and neural networks. The system uses trained classifiers to automatically detect and classify surface defects in display panels, substituting physical inspection methods with computational intelligence to achieve higher detection accuracy.
Solution Approach 2:
The patent employs synthetic data generation through data augmentation techniques, creating artificial defect samples that replicate real defect patterns. This allows the training dataset to include diverse defect variations without requiring extensive physical defect samples, thereby improving model generalization while managing system complexity.
2Measurement precision
If AI-based defect detection is implemented, then defect classification accuracy improves, but the requirement for training data increases and processing time extends
Solution Approach 1:
The patent performs data preprocessing and augmentation in advance to create a comprehensive training dataset before model training. Synthetic defect samples are generated beforehand to ensure the training data is ready, reducing the time required during actual defect detection operations.
Solution Approach 2:
The patent implements a dynamic training approach where the model can be retrained and updated as new defect patterns are discovered. The system adapts to new defect types by incorporating additional training samples, allowing the classification accuracy to improve over time without requiring complete system redesign.
3Reliability
If the training dataset is expanded with more defect samples, then model performance improves, but data collection and processing requirements increase
Solution Approach 1:
The patent uses data augmentation to create synthetic copies of defect samples through transformations such as rotation, scaling, and noise addition. This technique multiplies the effective training data volume from limited physical defect samples, improving model reliability without requiring proportional increases in physical defect collection.
Solution Approach 2:
The patent varies parameters of existing defect samples to generate diverse training data, including changes in defect size, position, orientation, and intensity. By manipulating these parameters computationally, the system generates a large volume of training samples from a small set of original defect images, reducing the need for extensive physical data collection.
Data Source
AI summary
Systems and methods for classifying products are disclosed. A first data sample having a first portion and a second portion is identified from a training dataset. A first mask is generated based on the first data sample, where the first mask is associated with the first portion of the first data sample. A second data sample is generated based on a noise input. The first mask is applied to the second data sample for outputting a third portion of the second data sample. The third portion of the second data sample is combined with the second portion of the first data sample for generating a first combined data sample. Confidence and classification of the first combined data sample are predicted. The first combined data sample is added to the training dataset in response to predicting the confidence and the classification.