Image Recognition Device Using Dynamic Ensemble Sections
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition devices face variability in accuracy due to differences in machine learning algorithms and learning data, leading to inconsistent performance in recognizing objects, and they often require significant resources for operation.
Innovation Solution
An image recognition device that combines the characteristics of differently learned neural network classifiers by detecting changes in frame data and applying these classifiers to sequential sections of data, sharing operation resources and alternately activating them to improve accuracy while reducing resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple differently learned neural network classifiers are applied to improve image recognition accuracy, then recognition accuracy is improved, but resource consumption increases
Solution Approach 1:
The patent divides the image recognition process into multiple ensemble sections, where each section applies a subset of neural network classifiers to specific frame data. This segmentation allows the system to distribute computational load across time and space, reducing peak resource consumption while maintaining high recognition accuracy through multiple classification attempts.
Solution Approach 2:
The patent implements periodic action by alternately activating different neural network classifiers across sequential frame data in ensemble sections. Instead of running all classifiers simultaneously on all data, the system cycles through different classifier combinations in different time periods, achieving high accuracy through repeated classification opportunities while controlling resource usage through temporal distribution.
2Measurement precision
If multiple differently learned neural network classifiers are applied to improve image recognition accuracy, then recognition accuracy is improved, but device complexity increases
Solution Approach 1:
The patent segments the complex system into manageable components: frame data change detector, ensemble section controller, image recognizer with multiple classifiers, and recognition result classifier. Each component has a specific function, making the overall complex system easier to manage and maintain while achieving high recognition accuracy through coordinated operation of these segmented parts.
Solution Approach 2:
The patent introduces dynamic control through the ensemble section controller, which dynamically determines how to divide frame data into ensemble sections based on frame data changes. This dynamic adaptation allows the system to adjust its complexity in real-time, applying multiple classifiers only when necessary to maintain accuracy while reducing complexity during stable periods.
3Measurement precision
If frame data is divided into multiple ensemble sections with different neural network classifiers, then recognition accuracy is improved, but processing time increases
Solution Approach 1:
The patent applies periodic action by processing different ensemble sections with different neural network classifiers in a cyclic manner. The ensemble section controller divides frame data into multiple sections and alternates which classifiers process which sections, enabling parallel processing opportunities and reducing overall processing time compared to sequential application of all classifiers to all data.
Solution Approach 2:
The patent uses dynamic adaptation to adjust processing time by monitoring frame data changes. When frame data remains relatively stable, the ensemble section controller can reduce the number of ensemble sections or reuse previous classification results, thereby reducing processing time while maintaining accuracy. When changes are detected, the system dynamically increases processing thoroughness.
Data Source
AI summary
Provided is an image recognition device. The image recognition device includes a frame data change detector that sequentially receives a plurality of frame data and detects a difference between two consecutive frame data, an ensemble section controller that sets an ensemble section in the plurality of frame data, based on the detected difference, an image recognizer that sequentially identifies classes respectively corresponding to a plurality of section frame data by applying different neural network classifiers to the plurality of section frame data in the ensemble section, and a recognition result classifier that sequentially identifies ensemble classes respectively corresponding to the plurality of section frame data by combining the classes in the ensemble section.