X-Ray Inspection Model Retraining From Corrected Inference Results
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The accuracy of inference results from trained models used in X-ray image analysis decreases due to changes in the types of components mounted on inspection objects over time, leading to a mismatch between training data and actual inspection data.
Innovation Solution
A method for recreating a trained model by acquiring and correcting inference results based on user input, storing these corrections, and using them to recreate the model, thereby reducing the burden of creating new training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a trained model is created using initial training data, then the model can perform inference on inspection objects, but the accuracy of inference results decreases when component types change over time
Solution Approach 1:
The system collects user corrections to inference results and feeds them back into the training data. The correction information storage unit stores corrected inference results, which are then used to retrain the model, creating a continuous feedback loop that improves accuracy over time while adapting to changing component types.
Solution Approach 2:
The system performs preliminary corrections to inference results before they are used for training. By correcting inference results in advance and storing them, the system prepares high-quality training data that reflects actual component variations, enabling the model to adapt proactively to changing conditions.
2Reliability
If new training data is collected separately to update the model, then the model accuracy can be improved, but the workload and time required for model maintenance increases
Solution Approach 1:
The system automatically collects and stores corrected inference results from normal inspection operations, eliminating the need for separate manual data collection efforts. The correction information storage unit captures training data automatically during routine use, allowing the model to be maintained without additional time investment from users.
Solution Approach 2:
The system continuously accumulates corrected inference results during normal inspection operations. Rather than requiring periodic separate data collection campaigns, the system continuously gathers training data from ongoing operations, making model maintenance an uninterrupted, integrated process.
3Adaptability or versatility
If a general-purpose trained model is used for multiple inspection object types, then the model can handle various components, but the accuracy decreases for specific component type changes
Solution Approach 1:
The system applies different correction strategies to different component types based on their specific characteristics. By storing and processing corrections separately for different inspection object types, the system tailors the training data to each component category, improving accuracy for specific components while maintaining general-purpose capability.
Data Source
AI summary
In this method for recreating a trained model, a first trained model 31a is a general-purpose trained model corresponding to a plurality of types of inspection objects 90. The method comprises a step of obtaining a first inference result 50a, obtained by inputting an X-ray image 40 into the trained model 31 for any X-ray image 40 of the plurality of types of inspection objects 90; a step of obtaining a first corrected inference result 51a by correcting the first inference result 50a; a step of storing the first corrected inference results 51a for the plurality of types of inspection objects 90 in association with their respective X-ray images 40; and a step of recreating the first trained model 31a using the associatedly stored X-ray images 40 and first corrected inference results 51a.