Image Analysis Device For Task-Specific Recipe Creation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image analyzing devices face challenges in creating a trained model suitable for a specific image analysis task, making it difficult to combine desired analysis processing freely and resulting in a high user workload.
Innovation Solution
An image analyzing device that includes an image holder, trained model registration, algorithm holder, recipe creation part, and analysis execution part, allowing users to create and combine trained models and algorithms to generate analysis recipes tailored to specific image analysis tasks, with automatic data set creation for machine learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If image analysis using machine learning is implemented, then automatic acquisition of parameters for identifying object regions is improved, but the workload for creating and managing trained models remains large
Solution Approach 1:
The system segments the trained model into multiple independent components: image acquisition algorithms, label image generation algorithms, and parameter extraction algorithms. Each component can be independently selected, modified, and combined through the recipe creation interface, allowing users to customize analysis pipelines without recreating entire models from scratch.
Solution Approach 2:
The analysis recipe creation interface provides universal functionality by allowing selection from multiple pre-stored trained models and algorithms. A single recipe framework can accommodate different image types, analysis purposes, and parameter combinations, making the system adaptable to various image analysis tasks without requiring separate custom development for each case.
2Adaptability or versatility
If multiple trained models are created for different analysis purposes, then analysis versatility is improved, but device complexity and user workload increase
Solution Approach 1:
Multiple trained models and algorithms are merged into a unified recipe creation interface. The system combines stored trained models with selectable analysis algorithms to generate integrated analysis pipelines. This merging allows users to access diverse analysis capabilities through a single standardized interface rather than managing separate complex model systems.
Solution Approach 2:
The system performs preliminary actions by pre-storing multiple trained models and analysis algorithms in accessible repositories before actual image analysis tasks. The recipe creation interface pre-configures combination options and parameters, so users can directly select and execute analysis pipelines without performing time-consuming model training or configuration during the analysis workflow.
3Measurement precision
If custom trained models are created for specific analysis tasks, then measurement precision is improved, but the time and effort required for model creation increases
Solution Approach 1:
The system uses copying by storing and reusing proven trained models and analysis algorithms that have been previously developed and validated. Instead of creating new models from scratch for each task, users can copy existing trained models and algorithms into new recipes, modifying parameters as needed. This preserves the precision of well-tested models while dramatically reducing creation time.
Data Source
AI summary
An image analyzing device (1) includes an image holder (8) that holds an image, a trained model registration part (10) configured to register trained models created by machine learning, a trained model holder (12) that holds the trained models registered by the trained model registration part (10), an algorithm holder (14) that holds a plurality of analysis algorithms for executing analysis processing of an image, a recipe creation part (18) configured, for an image to be analyzed optionally selected from among images held in the image holder (8), to create an analysis recipe for analyzing the image to be analyzed by combining a trained model selected from the trained models held in the trained model holder (10) and an analysis algorithm optionally selected from the plurality of analysis algorithms held in the algorithm holder (14), and an analysis execution part (20) configured to execute analysis of the image to be analyzed based on the analysis recipe created by the recipe creation part (18).