Image Classification Subclassification for Teaching Data Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image classification systems face challenges in creating appropriate teaching data, as non-classification target classes often contain images with multiple features, making it difficult to divide them effectively for each feature, leading to suboptimal classification accuracy and high man-hours required for manual division.
Innovation Solution
An image classification device and method that inputs image groups belonging to classification and non-classification target classes, extracts features, clusters them, and divides the non-classification target class into subclasses for each feature, thereby generating appropriate teaching data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a non-classification target class is created to collect all images other than classification target classes, then the classification process is simplified, but images with multiple features are mixed in a single class which reduces classification accuracy
Solution Approach 1:
The non-classification target class is automatically divided into multiple subclasses based on feature similarity. The subclassification unit extracts features from images in the non-classification target class and clusters them to create multiple subclasses, ensuring that images with similar features are grouped together. This segmentation resolves the contradiction by maintaining the simplified classification process while improving accuracy through feature-based subdivision of the non-classification target class.
2Manufacturing precision
If the non-classification target class is manually divided for each feature to achieve compact learning, then classification accuracy is improved, but the man-hours required for work increase significantly
Solution Approach 1:
The system performs automatic subclassification of the non-classification target class using feature extraction and clustering algorithms. The subclassification unit automatically extracts features from images and clusters them without requiring manual expert intervention. This self-service approach resolves the contradiction by achieving feature-based division that improves classification accuracy while eliminating the time-consuming manual work, as the system autonomously creates meaningful subclasses.
3Loss of time
If a generation model of the classification target class is used to automatically divide the non-classification target class, then manual work is reduced, but images having small similarity with all classification target classes cannot be properly divided
Solution Approach 1:
The patent replaces the generation model approach with a feature extraction and clustering-based subclassification system. Instead of relying on a generation model that may fail to capture diverse features, the system directly extracts features from images in the non-classification target class and uses clustering algorithms to group them. This substitution resolves the contradiction by providing a more robust automatic division method that can handle images with varying features, including those with small similarity to classification target classes, thereby improving feature-based division accuracy while maintaining low manual work requirements.
Data Source
AI summary
The objective of the present invention is to provide an image classification device and a method therefor with which suitable teaching data can be created. An image classification device that carries out image classification using images which are in a class to be classified and include teaching information, and images which are in a class not to be classified and to which teaching information has not been assigned, said image classification device being characterized by being provided with: an image group input unit for receiving inputs of an image group belonging to a class to be classified and an image group belonging to a class not to be classified; and a subclassification unit for extracting a feature amount for each image in an image group, clustering the feature amounts of the images in the image group belonging to a class not to be classified, and thereby dividing the images into sub-classes.