Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

An end-to-end based human body articulation point detection and classification method

A technology of human joints and classification methods, applied in neural learning methods, instruments, biological neural network models, etc., can solve problems such as weak robustness and unintuitive display of results, and achieve enhanced correlation and spatial position correlation , the effect of high precision estimation effect

Inactive Publication Date: 2018-12-07
UNIV OF ELECTRONICS SCI & TECH OF CHINA
View PDF5 Cites 28 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to: provide an end-to-end human body joint point detection and classification method, which solves the problem that the existing output method has weak robustness, and the display of the result is not intuitive during the training process

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
  • An end-to-end based human body articulation point detection and classification method
  • An end-to-end based human body articulation point detection and classification method
  • An end-to-end based human body articulation point detection and classification method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0034] The end-to-end human joint point detection and classification method includes the following steps in sequence:

[0035] S1. Construct a deep learning network based on stacked hourglass and conditional random field theory, and initialize the network;

[0036] S2. Train the network parameters, and obtain network weight parameters suitable for human body pose estimation through forward propagation;

[0037] S3. Input the preprocessed image into the trained hourglass neural network, obtain the heat map distribution of the corresponding joint points through forward propagation, and then use the conditional random field model to strengthen the spatial position relationship between each node, and finally Obtain the probability and statistics distribution of the appearance of each node, and connect the joint points with the output prediction results using preset rules, so as to realize end-to-end human pose estimation.

[0038] Considering adapting to joint point regression in...

Embodiment 2

[0060] The present invention uses 25,000 pictures in the MPII Human Pose Dataset as training and testing samples, and uses the center position of each picture as the estimated position of the body. And the preprocessing work of data augmentation such as simple scale scaling, rotation, adding noise, etc. is performed on the training pictures.

[0061] figure 1 It shows the basic flow chart of the deep learning network training and testing process based on stacked hourglass and conditional random field theory proposed by the present invention. Such as figure 1 As shown, in the training phase of the network, the image must first be preprocessed, and each video frame or image is cut into a size of 256x256 and then input into the network. For the training phase, first initialize the network parameters, set the initial learning rate to 0.0007, and adjust the number of human body nodes according to the needs, which is set to 16 here. In the forward propagation process, the paramet...

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 an end-to-end based human body articulation point detection and classification method. The realization method thereof comprises the steps of constructing a deep learning network and training and testing input picture data. The deep learning network includes a basic network layer structure, a pyramid pooling residual module and a stacked hourglass-type module. In a trainingphase: pre-processing operation of an input image and initialization of network parameters are carried out; the processed image is input into the deep learning network based on the stacked hourglass-type module and the conditional random field theory to undergo training; and network weight parameters are updated according to a Softmax loss function. In a test phase: forward-propagation of the testimage is performed by using the deep learning network model parameters obtained by the learning to obtain the probability distribution of the test articulation points, and the articulation points aresequentially connected by using known criteria to obtain the result image.

Description

technical field [0001] The invention relates to the fields of image segmentation, pattern recognition and computer vision, in particular to an end-to-end human body joint point detection and classification method. Background technique [0002] The core task of computer vision is to solve the two problems of detection and classification, and human body pose estimation, as one of the hot research in the field of modern intelligent recognition, not only has far-reaching research significance in the field of academic research, but also has a profound impact on the safety of our daily life. Detection, such as the detection of dangerous actions in streets with a lot of traffic and various public events, has profound practical significance. The primary goal of human pose estimation is to estimate the specific position of the key nodes of the body from a picture, and classify the nodes belonging to the same person from the multi-person scene. [0003] Since the development of neura...

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): G06K9/00G06N3/04G06N3/08
CPCG06N3/08G06V40/10G06N3/045
Inventor 程建林莉王艳旗苏炎洲白海伟
Owner UNIV OF ELECTRONICS SCI & TECH OF CHINA
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com