Presentation Information Generation for Targeted Training Data Collection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for collecting training data for machine learning are inefficient and costly, particularly when dealing with low-frequency data scenarios, and there is a need for more effective data collection strategies.
Innovation Solution
A presentation information generation apparatus and method that includes a data acquiring circuitry, a determining circuitry to assess data acquisition status based on comprehensive conditions, and an information generating circuitry to create presentation information for insufficient data, guiding users to collect missing data efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple cameras are used to reduce image capturing time, then productivity is improved, but device complexity and cost increase
Solution Approach 1:
The system enables a single camera to perform multiple capturing tasks by automatically selecting and executing different capturing strategies based on data sufficiency analysis. The camera serves itself by adapting its operation mode (sequential capturing, simultaneous capturing, or skipping) without requiring multiple physical cameras, thus maintaining productivity while reducing device complexity
Solution Approach 2:
The image capturing process is made dynamic by continuously monitoring data acquisition status and adjusting capturing parameters in real-time. The system dynamically switches between different capturing modes (sequential, simultaneous, skipping) based on the current state of data collection, allowing a single camera to achieve the productivity of multiple cameras through adaptive behavior
2Manufacturing precision
If detailed instructions are provided for each scene to ensure comprehensive data collection, then manufacturing precision is improved, but loss of time increases
Solution Approach 1:
The system applies partial action by selectively instructing the camera to capture only the specific scenes or data types that are currently insufficient, rather than requiring comprehensive capturing of all possible scenes. This reduces time loss while maintaining data collection completeness by focusing efforts only where needed
Solution Approach 2:
The system continuously monitors the acquisition status of training data and provides feedback-based instructions to the camera. By analyzing which data categories are sufficient and which are insufficient, the system generates targeted instructions that ensure comprehensive data collection without requiring time-consuming capturing of already-sufficient data
3Device complexity
If sequential image capturing is used to reduce cost, then device complexity is reduced, but productivity deteriorates
Solution Approach 1:
The system implements periodic action by alternating between sequential and simultaneous capturing modes based on data sufficiency requirements. When data collection is comprehensive, the system uses cost-effective sequential capturing; when data is insufficient, it switches to faster simultaneous capturing, thus maintaining productivity while using a single camera
Solution Approach 2:
The system applies skipping by allowing the camera to skip certain capturing steps when data sufficiency is already achieved for particular categories. This rushes through the data collection process efficiently by avoiding redundant sequential capturing operations, thereby improving productivity without increasing device complexity
Data Source
AI summary
A presentation information generation apparatus includes: a data acquiring circuitry that acquires data from a sensor; a determining circuitry that determines an acquisition status of acquisition target data based on a comprehensive condition and the data acquired, the comprehensive condition being a condition related to the acquisition target data; and an information generating circuitry that generates presentation information corresponding to insufficient data among pieces of the acquisition target data for which the acquisition status is inadequate.