Hierarchical Object Classification Model With Connected Classifiers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing hierarchical image classification methods face accuracy issues as the number of classes increases, leading to decreased performance.
Innovation Solution
A learning model with a main network that extracts feature amounts and a sub network that outputs hierarchical classes, where higher-level classifiers are connected to lower-level classifiers, utilizing convolution layers and fully-connected neural networks for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the number of classes in each layer increases to handle more complex product categories, then the classification coverage is improved, but the accuracy of hierarchical classification decreases
Solution Approach 1:
The patent divides the classification task into multiple hierarchical layers, where each layer handles a specific level of granularity. The main network extracts features at different levels, and multiple classifiers process these features for different class granularities. This segmentation allows the system to maintain high accuracy at each level while covering a broad range of product categories across the hierarchy.
Solution Approach 2:
The patent introduces a hierarchical dimension to the classification system, organizing classes from coarse to fine granularities across multiple levels. By adding this dimensional structure, the system can simultaneously handle both broad category coverage and precise classification within each category, resolving the trade-off between versatility and accuracy.
2Adaptability or versatility
If multiple classifiers are used to handle different class granularities, then the classification capability is improved, but the device complexity increases
Solution Approach 1:
The main network in the patent serves multiple functions by extracting features that are utilized by all classifiers at different hierarchical levels. This shared feature extraction mechanism allows the system to maintain high classification capability across multiple granularities while avoiding the need for separate feature extraction networks for each classifier, thereby controlling overall system complexity.
Solution Approach 2:
The patent merges the feature extraction functionality into a single main network that serves all classifiers, rather than having separate extraction networks for each classifier. This consolidation reduces redundant components and simplifies the overall architecture while maintaining the ability to handle multiple class granularities through the shared features.
Data Source
AI summary
An information processing apparatus includes an acquiring unit configured to acquire an object image that includes an object, and a predicting unit configured to predict a class of the object by applying the object image acquired by the acquiring unit to a learning model. The learning model includes a main network (201) that extracts feature amounts regarding hierarchical classes of the object based on the object image, and a sub network (202) that outputs hierarchical classes of the object based on the feature amounts. The main network (201) includes extractors that extract the feature amounts. The sub network (202) includes classifiers that output classes of the object based on the feature amounts, and in the sub network, a higher-level classifier has a connection to one or more lower-level classifiers.