Neural Network Defect Detection With Automatic Training Parameter Tuning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing deep learning-based defect detection systems in screen production require manual adjustment of training parameters, leading to resource wastage and potential human errors due to equipment, parameter, operation, and environmental interference.
Innovation Solution
A method for detecting defects using a neural network model that automatically configures training parameters based on feature information, including learning rate descent strategy, total number of training rounds, and test strategy, reducing the need for human intervention and enhancing accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual adjustment of training parameters is used in deep learning-based defect detection, then human resources can be flexibly allocated, but resource wastage and human errors occur
Solution Approach 1:
The system automatically adjusts training parameters based on dataset characteristics and model performance feedback, eliminating the need for manual human intervention. The model self-optimizes learning rates, training rounds, and test strategies through embedded algorithms that monitor training progress and autonomously modify parameters to achieve optimal detection accuracy while avoiding resource wastage from manual trial-and-error adjustments
2Ease of operation
If manual adjustment of training parameters is used, then flexibility in parameter tuning is achieved, but human errors and time consumption increase
Solution Approach 1:
The system pre-configures multiple parameter adjustment strategies and performance evaluation criteria before training begins. Based on initial dataset analysis, the system pre-determines optimal parameter ranges and adjustment sequences, allowing rapid adaptation during training without time-consuming manual analysis. This preliminary preparation enables the system to quickly respond to performance feedback and adjust parameters efficiently
Solution Approach 2:
The system continuously monitors model performance metrics during training and uses this feedback to automatically adjust training parameters. The feedback mechanism compares actual detection accuracy against target thresholds and dynamically modifies learning rates, training rounds, and test strategies accordingly, eliminating the need for time-consuming manual parameter tuning while maintaining operational flexibility
3Reliability
If automated parameter configuration is implemented, then human errors are reduced, but system complexity increases
Solution Approach 1:
The system introduces an intermediate parameter management layer that sits between the training data and the neural network model. This intermediary component automatically translates dataset characteristics into appropriate training parameters, shielding users from complex parameter adjustments while ensuring optimal model training. The intermediary handles the complexity of parameter configuration internally through automated algorithms, presenting a simplified interface to users
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
The present disclosure relates to the technical field of defect detection, and in particular, to a method and device for detecting a defect, method and device for training a model, and electronic device. The method includes: acquiring a sample data set including defective product data, and identifying feature information of the sample data set, the feature information including a number of samples in the sample data set; acquiring an initial model, the initial model being a neural network model; configuring a training parameter based on the feature information; obtaining a target model by training, according to the training parameter, the initial model with the sample data set; and obtaining defect information of a product corresponding to the sample data set by inputting real data of the product into the target model. The training parameter includes at least one of a learning rate descent strategy, a total number of training rounds and a test strategy, the learning rate descent strategy includes a number of learning rate descents and a round number when a learning rate descends, and the test strategy includes a number of tests and a round number when testing. The solution saves human resources.