Multi-Model Inference for High-Quality Training Data Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image analysis devices using machine learning require large amounts of training data, which are often of poor quality due to low accuracy in training data generation, necessitating significant manual effort.
Innovation Solution
A data analysis system comprising a central management device, generation device, and edge systems that utilize multiple inference models to generate high-quality training data by selecting optimal inference results and deploying learning models in edge environments for real-time data analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If multiple inference models are used to generate training data, then the quality of training data improves, but the device complexity increases
Solution Approach 1:
The generation device is divided into multiple independent inference models (first inference model, second inference model, etc.), each responsible for generating training data from different perspectives or using different algorithms. This segmentation allows each model to specialize in specific aspects of data generation, improving overall data quality while maintaining manageable complexity through modular architecture
Solution Approach 2:
The generation device is designed as a multi-functional system that can perform multiple inference operations simultaneously using different inference models. Each model serves a specific function in the training data generation process, and the device can selectively apply different models based on the requirements, making the system universally applicable to various data generation scenarios
2Manufacturing precision
If manual analysis is used to generate training data, then the quality of training data improves, but the productivity decreases
Solution Approach 1:
The system employs inference models that automatically analyze and generate training data without requiring manual intervention. The models self-service by processing input data, generating predictions, and producing training data autonomously, thereby maintaining high data quality through algorithmic precision while achieving high productivity through automated operation
Solution Approach 2:
Manual mechanical analysis operations are replaced with automated inference models that use computational algorithms to perform data analysis. This substitution eliminates the need for human analysts while maintaining or improving data quality through consistent, repeatable computational processes that can operate at high speeds
3Device complexity
If a single inference model is used, then the device complexity is reduced, but the measurement precision of training data generation deteriorates
Solution Approach 1:
Multiple inference models are merged into a single integrated generation device that operates cooperatively. The models combine their individual strengths to produce training data with higher accuracy than any single model could achieve alone, while the unified device structure manages complexity through coordinated operation of the merged components
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A generation device executes an inference process of inputting analysis target data to a plurality of inference models and outputting a plurality of inference results related to an object included in the analysis target data from the plurality of inference models, a determination process of determining, based on the plurality of inference results, a specific inference result from the plurality of inference results output by the inference process, and a generation process of generating a training data set including the specific inference result determined by the determination process and the analysis target data.