Multimodal Prediction Model for Concurrent Property Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods are inadequate for concurrently predicting multiple characteristics of an object, as they often focus on a single criterion, failing to meet the requirements for a socially valuable product that needs to satisfy a plurality of characteristics.
Innovation Solution
A prediction device and system that acquires both non-scientific information, such as images, and scientific information, such as spectra, to predict a plurality of characteristics using a trained discriminator and machine learning, enabling concurrent prediction of mechanical, physical, and functional properties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single characteristic is predicted using existing methods, then the prediction accuracy for that specific characteristic is improved, but the ability to concurrently predict multiple characteristics deteriorates
Solution Approach 1:
The patent combines multiple prediction models into a single integrated prediction device that can concurrently predict multiple characteristics. Each prediction model processes different characteristics (e.g., mechanical properties, physical properties, chemical properties) simultaneously using the same input data (image and spectrum), enabling concurrent multi-characteristic prediction while maintaining accuracy for each individual characteristic.
Solution Approach 2:
The prediction device is designed as a universal system that can predict multiple different characteristics using the same basic architecture. The device accepts image data and spectrum data as input and can output predictions for various characteristics (mechanical, physical, chemical, functional) depending on the trained prediction models, making it multi-functional and adaptable to different prediction tasks.
2Productivity
If multiple characteristics are predicted concurrently using a single system, then the productivity is improved, but the device complexity increases
Solution Approach 1:
The prediction device is segmented into multiple independent prediction models, each responsible for predicting a specific characteristic. Each model receives the same input data (image and spectrum) and processes it independently through its own trained parameters. This segmentation allows concurrent prediction of multiple characteristics while keeping each individual model relatively simple and manageable.
Solution Approach 2:
The patent introduces a new dimension to the prediction system by adding spectrum data as an additional input dimension alongside image data. This dimensional expansion enables the system to capture more comprehensive information about the object, improving the ability to predict multiple characteristics concurrently without significantly increasing the complexity of individual prediction models.
3Ease of operation
If only image data is used for prediction, then the ease of operation is improved, but the measurement precision for multiple characteristics deteriorates
Solution Approach 1:
The patent merges image data and spectrum data into a unified prediction approach. The system accepts both types of data as input and processes them together through the prediction models. This combination allows the system to maintain ease of operation (by providing comprehensive data) while significantly improving measurement precision for multiple characteristics through the synergistic information from both image and spectrum sources.
Solution Approach 2:
The prediction approach treats the combination of image data and spectrum data as a composite information structure. Just as composite materials combine different material properties to achieve superior performance, the system combines different data types (visual and spectral information) to achieve superior prediction accuracy for multiple characteristics simultaneously.
Data Source
AI summary
A prediction device includes an acquirer that acquires first information including an image regarding an object and second information including at least one of a character, a number, a chemical structure, and a spectrum regarding the object, and a predictor that predicts a plurality of characteristics of the object based on the acquired first information and the acquired second information.