Person Attribute Detection Using Multi-Region Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing attribute determining methods face challenges in detecting and determining a person's attributes when characteristic facial parts are hidden, and require the person to face nearly frontwards, making it difficult to implement in various applications due to irregular human movements and the need for specific camera placements.
Innovation Solution
An attribute determining method and system that acquires images and detects at least two attribute determination regions, including head, facial, and whole-body regions, to determine attributes based on these regions, even when characteristic facial parts are hidden, using a combination of image acquisition, region detection, and attribute determination models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If attribute determination is performed using only facial parts, then detection accuracy is improved when the face is visible, but the system cannot determine attributes when characteristic facial parts are hidden or when the person is not facing front
Solution Approach 1:
The system segments the attribute determination process into multiple independent region detectors, each specialized for detecting specific body parts (face, head, shoulders, upper body). This segmentation allows the system to flexibly combine results from different regions depending on which are visible, resolving the contradiction between accuracy and adaptability.
Solution Approach 2:
The system creates a universal attribute determination framework that can process multiple types of input regions (face, head, shoulders, upper body) through a common attribute determination model. This multi-functionality enables the system to adapt to various camera angles and occlusion conditions while maintaining consistent attribute determination capabilities.
2Measurement precision
If a camera is placed to capture front-facing persons, then attribute determination accuracy is improved, but the system becomes difficult to implement in general surveillance scenarios
Solution Approach 1:
The system applies local quality by training different region detection models optimized for specific body parts and their characteristic appearances. Each detector is specialized for its target region, allowing accurate detection even when the overall pose varies, thus enabling accurate attribute determination without requiring strict front-facing camera placement.
3Measurement precision
If multiple attribute determination regions are detected and combined, then detection accuracy is improved for various orientations, but the system complexity increases
Solution Approach 1:
The system performs preliminary detection of multiple attribute determination regions (face, head, shoulders, upper body) before the final attribute determination step. This preliminary action prepares the input data in advance, allowing the subsequent attribute determination model to process pre-processed regional information, which improves accuracy while managing complexity through staged processing.
Data Source
AI summary
The present invention is to provide an attribute determining method, an attribute determining apparatus, a program, a recording medium, and an attribute determining system of high detection accuracy of a person with which an attribute of a person can be determined, for example, even in the case where characteristic parts of the face are hidden.The attribute determining method of the present invention includes an image acquiring step (S11) of acquiring an image of a person to be determined, an attribute determination region detecting step (S21) of detecting at least two attribute determination regions selected from the group consisting of a head region, a facial region, and other regions from the image of a person to be determined, and an attribute determining step (S22) of determining an attribute based on images of the at least two attribute determination regions.