Human posture estimation algorithm based on keypoint error weighted loss function

Jingying Xu; Yu Wang

doi:10.1117/12.3033785

13 June 2024 Human posture estimation algorithm based on keypoint error weighted loss function

Jingying Xu, Yu Wang

Proceedings Volume 13180, International Conference on Image, Signal Processing, and Pattern Recognition (ISPP 2024); 131803A (2024) https://doi.org/10.1117/12.3033785
Event: International Conference on Image, Signal Processing, and Pattern Recognition (ISPP 2024), 2024, Guangzhou, China

Abstract

The different scales and flexibility of human keypoints lead to smaller detection errors at facial keypoints and larger detection errors at keypoints on the limbs. To solve the above problem, this study takes Disentangled Keypoint Regression network as the baseline, and proposes to optimize the classification loss function weighted by using the keypoint detection error factor, and allocates more arithmetic resources to the keypoints with larger detection errors, to solve the imbalance in the detection errors among keypoints. Problem. At the same time, the design of the Recall Disentangled Keypoint Regression network is based on the idea of backtracking connection fusion of multi-stage feature information, compared with the DEKR network, each stage of the feature output node will backtrack to the feature information output from the previous network, to enhance the circulation of multi-stage information in the network, and to make full use of the network feature information to solve the problem of the difficulty in detecting the keypoints of the small-scale human. The method of this paper is validated on COCO2017 and CrowdPose, respectively, and the accuracy AP is improved by 1.2 and 1.1, respectively, while the human pose estimation is visualized on the images of the real scene, which verifies the effectiveness and practicality of the method of this paper.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Jingying Xu and Yu Wang "Human posture estimation algorithm based on keypoint error weighted loss function", Proc. SPIE 13180, International Conference on Image, Signal Processing, and Pattern Recognition (ISPP 2024), 131803A (13 June 2024); https://doi.org/10.1117/12.3033785

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
7 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Pose estimation

Error analysis

Visualization

Image resolution

Autoregressive models

Deep learning

Image classification

Show All Keywords

Keywords/Phrases

Search In:

Publication Years