Hierarchical AI Training Dataset With Acquisition Context Layers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI network training datasets lack effective representation of acquisition context and conditions, making it difficult to train AI networks that can adapt to varying scenarios, leading to suboptimal recognition performance.
Innovation Solution
A hierarchical dataset generation method that includes acquiring and storing vehicle data, sensor data, and context information, with context information generated through synthetic analysis and stored hierarchically, allowing for detailed description of acquisition conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If existing training datasets are configured only by describing GT descriptor on input sensor information, then the dataset structure is simple, but it is impossible to distinguish acquisition condition and context information, making it difficult to train AI effectively
Solution Approach 1:
The dataset is segmented into multiple hierarchical levels: sensor information, GT descriptor, acquisition condition, and context information. Each level captures different aspects of the data, allowing context information to be preserved without creating a monolithic complex structure. The segmentation enables selective access and processing of specific information types.
Solution Approach 2:
The patent adds a new dimensional layer to the dataset structure by introducing context information as a separate dimension alongside the traditional sensor-GT pairing. This dimensional expansion allows the system to preserve and utilize context information without fundamentally complicating the core data structure, as context operates in an additional informational dimension.
2Adaptability or versatility
If AI network is trained without distinguished context information, then the training process is simple, but the AI network cannot adapt to varying scenarios and has suboptimal recognition performance
Solution Approach 1:
Context information is extracted and organized in advance during the data acquisition and preprocessing phase, before the AI training begins. This preliminary organization of context data allows the AI network to access structured contextual information during training without adding complexity to the training algorithm itself, as the contextual structure is already prepared.
Solution Approach 2:
The patent introduces context information as an intermediary layer between the raw sensor data and the AI network processing. This intermediary structure organizes and presents contextual information in a standardized format, allowing the AI network to adapt to varying scenarios without directly processing raw, unstructured contextual data, thus managing complexity through intermediate organization.
3Measurement precision
If context information is hierarchically described on GT descriptor, then AI network training is improved with better recognition performance, but the data processing complexity increases
Solution Approach 1:
Context information is segmented into hierarchical categories (e.g., environmental context, temporal context, spatial context) and associated with GT descriptors at appropriate levels. This segmentation allows precise contextual information to be provided to the AI network for improved recognition, while the modular hierarchical structure manages processing complexity by organizing data in manageable segments rather than a single complex structure.
Data Source
AI summary
Provided are a method and a system for generating an AI training hierarchical dataset including data acquisition context information. A GT dataset generation method according to an embodiment of the present disclosure includes: acquiring and storing vehicle data; acquiring and storing sensor data generated at a sensor installed in a vehicle; and generating and storing context information which is information regarding a context at a time when the data is acquired. Accordingly, in generating a GT descriptor, various contexts, conditions at the time when data is acquired may be made to be easily analyzed, classified on the GT descriptor through a hierarchical dataset, which hierarchically describes context information at the time when sensor data is acquired on the descriptor, so that an AI network is effectively trained, and eventually, has high recognition performance.