Two-dimensional to three-dimensional human body posture estimation method

A technology of human body posture and three-dimensional posture, applied in the field of computer vision and pattern recognition, can solve the problems of indistinguishable and unfavorable mutual learning of different postures, and achieve the effect of accurate three-dimensional posture estimation, promotion of mutual learning, and accurate three-dimensional posture.

Pending Publication Date: 2021-01-15
BEIJING UNIV OF TECH
View PDF0 Cites 6 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

If only the global pose features are extracted, it will make it difficult to distinguish poses with different levels of human body parts, and it is not conducive to mutual learning between different poses.

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Two-dimensional to three-dimensional human body posture estimation method
  • Two-dimensional to three-dimensional human body posture estimation method
  • Two-dimensional to three-dimensional human body posture estimation method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0017] Such as Image 6 As shown, this two-dimensional to three-dimensional human pose estimation method, the method includes the following steps:

[0018] (1) Layered graph convolutional network: including feature enhancement module and layered graph convolution module; feature enhancement module consists of 3 layers of fully connected layers, of which the last two layers form a residual block, which is obtained from two-dimensional coordinates through the network The potential relationship between joint coordinates is excavated in order to enhance the feature representation of the human body; the layered graph convolution module has 6 layers, each layer corresponds to the division of the human body model under the current granularity, and is extracted through a unified feature extraction network block The characteristics of different granularities of the human body, and through hierarchical fusion, to obtain better characteristics;

[0019] (2) Feature extraction network bl...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a two-dimensional to three-dimensional human body posture estimation method, which can overcome the defects of three-dimensional labeling data and the influence of conditions such as background, lamplight, clothing shape, texture and skin color, extract features from multiple scales, promote mutual learning among different postures and obtain more accurate three-dimensionalpostures. The method comprises the following steps: (1) arranging a hierarchical graph convolutional network; (2) extracting network blocks in combination with features of a convolution layer and a non-local layer of a diagonal dominance graph; and (3) setting human body geometric constraints.

Description

technical field [0001] The invention relates to the technical field of computer vision and pattern recognition, in particular to a two-dimensional to three-dimensional human posture estimation method. Background technique [0002] The 3D human pose estimation based on computer vision technology has been widely used in many fields of human life, such as computer animation, medicine, human-computer interaction, behavior recognition and other fields. Based on the rapid development of neural network technology, estimating 3D human pose from RGB images not only eliminates the dependence on RGB-D sensors (such as Kinect), but also obtains a significant improvement in performance, which has become a current research hotspot. [0003] The existing image-based 3D human pose estimation is mainly divided into two categories: 1) directly estimate the 3D human pose from the image; 2) estimate the 2D pose from the image first, and then regress the 3D pose. The former is limited by limite...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
IPC IPC(8): G06K9/00G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06V40/23G06V40/103G06N3/047G06N3/045G06F18/253
Inventor 孔德慧吴永鹏王少帆李敬华王立春
Owner BEIJING UNIV OF TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap