Image Processor Classifier Generator Reduces Processing Load
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processors face increased processing load and time due to the need for multiple classifiers to detect specific targets like humans and pets, as general-purpose detectors often fail to detect certain faces, requiring dedicated classifiers which escalate the number of processing units and time.
Innovation Solution
An image processor configuration that includes a general-purpose classifier and dedicated classifiers, with a classifier generator that creates a registered image classifier by replacing weak classifiers from dedicated classifiers with those from the general-purpose classifier, optimizing detection processing by reducing the number of classifiers needed for targets detectable by the general-purpose classifier.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple dedicated classifiers are used to detect specific targets, then detection accuracy is improved, but processing load and time increase
Solution Approach 1:
The detection system is segmented into general-purpose classifiers for common targets and dedicated classifiers for specific targets. The image extracting part segments the input image to extract only relevant regions containing registered targets, reducing the processing area for dedicated classifiers and improving overall processing efficiency while maintaining detection accuracy.
Solution Approach 2:
The system performs partial action by using the general-purpose classifier first to detect common targets, then only applying dedicated classifiers to extract and process regions containing registered targets. This avoids excessive processing of the entire image with all classifiers, reducing processing load while maintaining detection accuracy for both general and specific targets.
2Measurement precision
If multiple dedicated classifiers are used to detect specific targets, then detection accuracy is improved, but device complexity increases
Solution Approach 1:
The general-purpose classifier serves multiple functions by detecting both common targets and acting as a pre-processing step for dedicated classifiers. The image extracting part also serves multiple functions by both segmenting the image and identifying regions for dedicated classifier processing. This multi-functionality reduces the need for separate dedicated classifiers for every target type, reducing device complexity while maintaining detection accuracy.
Solution Approach 2:
The image extracting part acts as an intermediary between the general-purpose classifier and dedicated classifiers. It receives the input image, extracts regions containing registered targets, and provides these extracted regions to the dedicated classifiers. This intermediary role allows the system to use fewer dedicated classifiers by pre-processing and filtering the input, reducing device complexity while maintaining detection accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In processing of detecting a target included in a registered image from a certain image, a processing load or a processing time is reduced. A registered image classifier is generated in order to detect the target included in the registered image by replacing some weak classifiers included in a dedicated classifier having the highest score of a detection result for the registered image with the weak classifiers included in a general-purpose classifier.