Learned model re-creating method, learned model re-creating system, and learned model re-creating program

By re-creating trained models with user-corrected inference results, the method addresses accuracy drops due to changing inspection object types, ensuring high accuracy and reducing user data preparation efforts.

JP2026031250APending Publication Date: 2026-02-24SHIMADZU SEISAKUSHO LTD
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
JP2024134657
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2024-08-09
Publication Date
2026-02-24

AI Technical Summary

Technical Problem

The accuracy of inference results using trained models for analyzing X-ray images decreases due to changes in the types of components mounted on inspection objects over time, leading to discrepancies between training data and actual inspection data.

Method used

A method for re-creating trained models by acquiring and correcting inference results from user inputs, storing these corrections, and using them to recreate the model, thereby reducing the need for separate training data creation and improving model suitability for evolving inspection objects.

Benefits of technology

The recreated trained models maintain high accuracy by incorporating user-corrected inference results, adapting to changes in inspection object types and reducing user burden in data preparation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2026031250000001_ABST
    Figure 2026031250000001_ABST
Patent Text Reader

Abstract

To suppress a decrease in accuracy of an inference result of an inspection object by a learned model even when a tendency of a type of a component or the like mounted on the inspection object changes.SOLUTION: In the learned model re-creation method, the first learned model 31a is a general-purpose learned model corresponding to a plurality of types of examination objects 90, and the learned model re-creation method includes a step of acquiring a first inference result 50a obtained by inputting the X-ray image 40 to the learned model 31 for any X-ray image 40 of the plurality of types of examination objects 90, a step of acquiring a first corrected inference result 50a obtained by correcting the first inference result 51a, a step of storing the first corrected inference result 51a for the plurality of types of examination objects 90 in association with each X-ray image 40, and a step of re-creating the first learned model 31a using the X-ray image 40 and the first corrected inference result 51a stored in association with each other.SELECTED DRAWING: Figure 16
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • X-ray imaging system and method for generating trained model

    JP2024029975A