Recognition Model Distribution System for Unknown Scene Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image recognition technologies for automatic driving, such as those described in Patent Literature 1, fail to effectively handle recognition errors in unknown scenes, leading to safety concerns in environments like intersections, highways, and school zones, as they do not provide feedback for unique events or unknown objects.
Innovation Solution
A recognition model distribution system that includes a data analysis unit to identify recognition failures, a parameter generation unit to create three-dimensional computer graphics videos, a three-dimensional object generation unit for traffic simulations, a teacher data generation unit to create learning data, and a recognition model distribution unit to update recognition models for improved accuracy in unknown scenes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning technology is applied for image recognition in automatic driving, then recognition performance for known scenes is improved, but recognition accuracy for unknown scenes or unique events cannot be ensured
Solution Approach 1:
The system performs preliminary actions by generating three-dimensional computer graphics videos of unknown or error-prone scenes before actual deployment. These synthetic videos are created in advance and used to train the recognition model, enabling it to learn from simulated scenarios that match real-world failure cases. This preliminary training with synthesized data prepares the model to handle unknown scenes it has never encountered in real driving data.
Solution Approach 2:
The system creates copies of problematic scenes through three-dimensional computer graphics rendering. By replicating unknown scenes, unique events, and recognition failure points in virtual form, the system generates synthetic training data that mirrors real-world conditions without requiring actual physical samples. These virtual copies enable the model to learn from scenarios that would be difficult or impossible to capture in real driving data.
2Measurement precision
If three-dimensional computer graphics videos are generated for training, then recognition accuracy for unknown scenes is improved, but system complexity increases
Solution Approach 1:
The three-dimensional computer graphics generation system serves multiple functions: it generates training data for unknown scenes, reproduces recognition failure points, creates diverse weather and lighting conditions, and synthesizes unique regional events. This single multi-functional system replaces what would otherwise require multiple separate data collection and processing systems, managing complexity through consolidation while delivering comprehensive training data coverage.
Data Source
AI summary
The purpose of the present invention is to provide a technology for updating a recognition model so that even if there were errors in recognition of an unknown scene or the like, the scene can be recognized quickly. The present invention is provided with a data analysis unit 11 that, on the basis of data from an outside recognition unit 32 provided to a vehicle, acquires from among previously stored recognition models a model approximate to a recognition model recognized by the outside recognition unit 32, and that reproduces the acquired model in the form of computer graphics images. The data analysis unit 11 is provided with: a difference extraction unit 114 that compares the reproduced computer graphics images and data from the outside recognition unit 32 and extracts a difference therebetween; an object recognition unit 116 that recognizes an object relating to the difference extracted by the difference extraction unit 114; and a scene reconfiguration unit 117 that creates computer graphics images having the object recognized by the recognition unit 116 reflected therein.