Recognition Model Integration Across Field Environments
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recognition systems fail to effectively utilize the characteristics of multiple learned discriminators across various field environments, leading to suboptimal recognition accuracy.
Innovation Solution
A recognition system comprising a server device and terminal devices that integrate and update models using a server-side model integration and update unit, allowing for the generation of high-accuracy models by optimizing the characteristics of multiple learned models from different environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the learned discriminator with the highest accuracy rate is selected and distributed to terminal devices, then the recognition accuracy is improved, but the characteristics of other learned discriminators are not effectively utilized
Solution Approach 1:
The patent merges multiple learned discriminators by integrating their model parameters (weights and biases) through averaging or weighted averaging. This combines the characteristics of all learned discriminators rather than selecting only one, thereby preserving valuable environment-specific features while improving overall recognition accuracy through ensemble integration.
2Measurement precision
If multiple learned discriminators are integrated to utilize their characteristics, then the model accuracy is improved, but the system complexity increases
Solution Approach 1:
The server device performs multiple functions: collecting model parameters from multiple terminal devices, integrating these parameters to generate improved model parameters, and distributing the updated parameters back to terminal devices. This multi-functional approach consolidates complexity in the server while keeping terminal devices simple, resolving the contradiction between accuracy improvement and system complexity.
Data Source
AI summary
The server device receives a model information from a plurality of terminal devices, and generates an integrated model by integrating the model information received from the plurality of terminal devices. The server device generates an updated model by learning a model defined by the model information received from the terminal device of update-target using the integrated model. Then, the server device transmits the model information of the updated model to the terminal device. Thereafter, the terminal device executes recognition processing using updated model.