Human Body Identification Using Key Point Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current human body identification methods in computer vision and deep learning face challenges in accurately detecting and classifying human body key points, leading to suboptimal identification accuracy and recall rates, especially in complex scenarios.
Innovation Solution
A method involving a human body detection model to identify preselected detection boxes, followed by key point detection and classification using a human body key point detection model and classification model, respectively, with the use of topological information to improve accuracy and recall rates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a human body detection model is trained mainly based on a human body detection box and a pedestrian detection set, then the detection process is simple, but the identification accuracy and recall rates are suboptimal
Solution Approach 1:
The patent segments the human body detection task into multiple stages: initial detection box generation, key point detection, key point classification, and final identification. This multi-stage segmentation allows each component to specialize in specific features, improving overall accuracy while managing complexity through modular processing
Solution Approach 2:
The patent introduces key point information as an additional dimension beyond traditional detection boxes. By detecting and classifying key points (body parts) in addition to bounding boxes, the system gains richer feature representation, improving identification accuracy through multi-dimensional feature fusion
2Reliability
If key point detection is performed on all preselected detection boxes, then more key point information is obtained, but the computational cost increases
Solution Approach 1:
The patent applies partial action by performing key point detection selectively rather than uniformly on all detection boxes. The key point classification model filters and prioritizes detection boxes, applying detailed key point analysis only to promising candidates, thus achieving high recall with reduced computational overhead
Solution Approach 2:
The patent performs preliminary filtering using the key point classification model before conducting full key point detection. This preliminary action identifies high-probability human body detection boxes, allowing subsequent detailed key point analysis to focus computational resources on the most relevant cases
Data Source
AI summary
A human body identification method, an electronic device and a storage medium, related to the technical field of artificial intelligence such as computer vision and deep learning, are provided. The method includes: inputting an image to be identified into a human body detection model, to obtain a plurality of preselected detection boxes; identifying a plurality of key points from each of the preselected detection boxes respectively according to a human body key point detection model, and obtaining a key point score of each of the key points; determining a target detection box from each of the preselected detection boxes, according to a number of the key points whose key point scores meet a key point threshold; and inputting the target detection box into a human body key point classification model, to obtain a human body identification result for the image to be identified.