AI Model Training Using Real-Aligned Simulation Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI models trained with a combination of small amounts of real data and large amounts of simulation data suffer from low accuracy due to the disparity between the simulation and real environments, making it difficult to collect sufficient real data for effective training.
Innovation Solution
A method involving a cloud service platform that adjusts simulation data based on real data to create more accurate simulation data, followed by model training using both real and adjusted simulation data to enhance the AI model's accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If a large amount of simulation data is generated to compensate for insufficient real data, then the quantity of training data is improved, but the accuracy of the AI model deteriorates due to disparity between simulation and real environments
Solution Approach 1:
The patent applies parameter changes by adjusting the simulation data through a transformation model that modifies parameters such as image resolution, color distribution, and other visual characteristics. This transforms the simulation data from its original state to a state that better matches real-world conditions, thereby improving model accuracy while maintaining the advantage of having large quantities of training data.
Solution Approach 2:
The patent introduces an intermediary transformation model that acts as a bridge between simulation data and real data. This intermediary component adjusts the simulation data to align with real-world characteristics, enabling the model to effectively utilize both simulation and real data without the accuracy degradation that would occur from using raw simulation data directly.
2Measurement precision
If real data is collected from the real environment to train the AI model, then the accuracy of the AI model is improved, but the quantity of available training data deteriorates due to difficulty in collecting sufficient real data
Solution Approach 1:
The patent applies copying by creating synthetic simulation data that replicates real-world scenarios. Instead of collecting limited real data, the system generates extensive simulation data that copies the essential characteristics of real environments, thereby obtaining large quantities of training data while maintaining accuracy through the copying of real-world patterns.
Solution Approach 2:
The patent uses parameter changes to transform simulation data to better match real data characteristics. By adjusting parameters such as visual properties and environmental conditions in the simulation data, the system creates a large quantity of training data that maintains high accuracy comparable to real data.
3Quantity of substance
If simulation data is used to train the AI model, then the quantity of training data is improved, but the reliability of the training process deteriorates due to environment disparity
Solution Approach 1:
The patent introduces an intermediary transformation model that mediates between simulation data and the training process. This intermediary component adjusts the simulation data to reduce environment disparity, thereby improving the reliability of the training process while maintaining the ability to use large quantities of simulation data.
Solution Approach 2:
The patent implements feedback mechanisms where the transformation model is iteratively improved based on the performance of the AI model and the quality of training outcomes. This feedback loop ensures that the simulation data adjustment process continuously improves reliability while maintaining sufficient training data quantity.
Data Source
AI summary
A model training method and apparatus, and a storage medium are provided, and pertain to the computer field. The method includes: obtaining a plurality of pieces of real data and a plurality of pieces of first simulation data, where the plurality of pieces of real data are data describing a real environment, the real environment is an environment to which a to-be-trained first artificial intelligence AI model is applied, the plurality of pieces of first simulation data are data describing a simulation environment, and the simulation environment is used to simulate the real environment; adjusting the plurality of pieces of first simulation data based on the plurality of pieces of real data to obtain a plurality of pieces of second simulation data; and performing model training based on the plurality of pieces of real data and the plurality of pieces of second simulation data to obtain the first AI model.