Human Attribute Recognition Using Confidence-Based Feature Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing human attribute recognition techniques face accuracy issues due to sparse feature spaces, particularly when high-dimensional features like Histogram of Oriented Gradient (HOG) and Local Binary Pattern (LBP) features are used, leading to fewer similar images being found for recognition, which affects the accuracy of human attribute recognition.
Innovation Solution
The proposed solution involves a recognition apparatus that extracts features from images and uses confidence information to determine sample information, reducing dimensionality and increasing the density of the feature space for more accurate attribute recognition. This apparatus includes a feature extraction unit, a feature obtainment unit, a sample information determination unit, and an attribute recognition unit, utilizing pre-determined sample information and confidence-based features to improve recognition accuracy and speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-dimensional features (HOG, LBP) are used for feature matching, then feature representation capability is improved, but the feature space becomes sparser and fewer similar images can be found
Solution Approach 1:
The patent segments the feature space by dividing similar images into multiple groups based on their feature similarities. Instead of treating all similar images as a single group, the system creates multiple sub-groups with different degrees of similarity to the input image. This segmentation allows the system to maintain high-dimensional feature representation while finding sufficient similar images by searching across multiple segmented groups rather than relying on a single sparse feature space.
2Measurement precision
If high-dimensional features are used for feature matching, then feature discrimination capability is improved, but recognition accuracy deteriorates due to sparser feature space
Solution Approach 1:
The patent introduces a new dimension to the feature space by adding group identifiers to the traditional high-dimensional features. Each similar image is assigned to a specific group based on feature similarity, creating an extended feature vector that includes both the original high-dimensional features and the group classification information. This dimensional extension allows the system to maintain the discrimination power of high-dimensional features while overcoming sparsity by leveraging group-based organization.
Solution Approach 2:
The patent introduces group information as an intermediary between the high-dimensional features and the recognition process. Instead of directly using high-dimensional features for recognition, the system first groups similar images based on feature similarities, then uses these groupings as an intermediate step to improve recognition. The group information acts as a mediator that bridges the gap between high-dimensional feature representation and reliable recognition by providing additional structural context.
Data Source
AI summary
A recognition apparatus for extracting first features of an object from a region of the object; obtaining second features of the object at least based on confidence information for first attribute of the object; determining sample information from pre-determined sample information based on the obtained second features; and recognizing the first attribute of the object based on the determined sample information.