Inspection System Machine Learning Model Versioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing inspection systems face challenges in accurately identifying inspection targets carrying handguns due to inconsistent determination results between different inspection devices, leading to difficulties in reliably identifying the presence and location of metal objects.
Innovation Solution
An inspection system comprising multiple devices, each using machine learning models to determine the presence and location of objects, with a processing unit generating update data to refine the models based on subsequent determinations, ensuring consistent and accurate identification across devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If multiple inspection devices are used to perform inspection in stages, then the quantity and detail of measurement information increases, but the consistency and reliability of determination results deteriorate due to inconsistent outputs from different devices
Solution Approach 1:
The system changes the parameter of machine learning model versions across different inspection devices. A first inspection device uses an older version of the machine learning model, while a second inspection device uses a newer version. This parameter change allows the system to utilize multiple devices with different measurement capabilities while maintaining result consistency through centralized model management and update propagation.
Solution Approach 2:
The system implements feedback by having the control device receive determination results from multiple inspection devices, generate update data based on these results, and transmit updated machine learning models back to the inspection devices. This closed-loop feedback mechanism ensures that all devices progressively align their determination capabilities, resolving the consistency issue while maintaining the benefits of multiple devices.
2Adaptability or versatility
If different machine learning model versions are used across inspection devices, then device complexity and adaptability increase, but determining which device has higher reliability becomes difficult
Solution Approach 1:
The control device performs preliminary actions by generating update data and transmitting updated machine learning models to inspection devices before new inspections are conducted. This ensures that all devices are proactively updated with the latest model versions, allowing the system to maintain high adaptability while ensuring that the most recent device (with the newest model) has the highest reliability for current determination tasks.
Solution Approach 2:
The system dynamically manages machine learning model versions across inspection devices. Instead of using static, identical models, the system allows different devices to operate with different model versions temporarily, then dynamically updates them through centralized control. This dynamic approach enables the system to adapt to new inspection challenges while maintaining the ability to determine which device has the highest reliability based on model version recency.
Data Source
AI summary
According to one embodiment, an inspection system includes a first measurement unit for measuring a target, a first determination unit for making a first determination on whether the target includes a predetermined object using a first machine learning model based on a measurement result by the first measurement unit, a second measurement unit for measuring the target, a second determination unit for making a second determination on whether the target includes the predetermined object based on a measurement result by the second measurement unit, and a processing unit for generating first update data of the first machine learning model based on the second determination.