Yang Chen1, Yaqian Wang1, Da Chen2, Jun Gu1,*
1Chongqing College of International Business and Economics, Chongqing 402582, China.
2Key Laboratory of Industrial Internet of Things and Network Control, Ministry of Education, Chongqing University of Posts and Telecommunications, Chongqing 400065, China.
*Corresponding author: Jun Gu
Abstract
In recent years, the study of human posture estimation based on deep learning has become one of the hottest research directions in the field of computer vision, which has very great research value. To address the challenges of excessive network parameters and high computational complexity in human pose estimation networks, we propose a lightweight human pose estimation network, named Pose, which is inspired by the Lightweight OpenPose architecture. Specifically, we introduce an enhanced GhostNet network for efficient feature extraction. Under identical image resolution and environmental configurations, experimental results on the COCO validation set demonstrate that Pose achieves a 6.7% reduction in parameter count and a 22.2% decrease in computational complexity compared to Lightweight OpenPose. These findings indicate that Pose not only maintains competitive performance in human pose estimation but also significantly reduces both model size and computational demands when compared to conventional networks such as OpenPose and Lightweight OpenPose.
References
[1] El Kaid A, Baina K. A Systematic Review of Recent Deep Learning Approaches for 3D Human Pose Estimation. J Imaging. 2023;9(12).
[2] Palermo M, Moccia S, Migliorelli L. Real-Time Human Pose Estimation on a Smart Walker using Convolutional Neural Networks. 2021.
[3] Dibenedetto G, Sotiropoulos S, Polignano M. Comparing Human Pose Estimation through deep learning approaches: An overview. Comput Vis Image Underst. 2025;252.
[4] Quero CO, Durini D, Rangel-Magdaleno J. Enhancing 3D human pose estimation with NIR single-pixel imaging and time-of-flight technology: a deep learning approach. J Opt Soc Am A Opt Image Sci Vis. 2024;41(3):10.
[5] Zhang W, Fang J, Wang X. Efficient Pose: Efficient human pose estimation with neural architecture search. 2021;7(3):335-347.
[6] Mo J, Jiang G, Yuan H. Adaptive Target Region Attention Network-based Human Pose Estimation in Smart Classroom. Int J Adv Comput Sci Appl. 2024;15(4).
[7] Toshev A, Szegedy C. DeepPose: Human Pose Estimation via Deep Neural Networks. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. United States: Google; 2014. p. 1653-1660.
[8] Wei SE, Ramakrishna V, Kanade T, Sheikh Y. Convolutional Pose Machines. In: Proceedings—29th IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, NV, USA; 2016. p. 4724-4732.
[9] Cao Z, Simon T, Wei SE, Sheikh Y. Realtime multi-person 2D pose estimation using part affinity fields. In: 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI, USA; 2017. p. 1302-1310.
[10] Osokin D. Real-time 2D Multi-Person Pose Estimation on CPU: Lightweight OpenPose. In: ICPRAM 2019 - Proceedings of the 8th International Conference on Pattern Recognition Applications and Methods. 2019. p. 744-748.
[11] Paoletti ME, Haut JM, Pereira NS. Ghostnet for hyperspectral image classification. IEEE Trans Geosci Remote Sens. 2021;59(12):10378-10393.
How to cite this paper
Improvement of a Human Pose Estimation Strategy
How to cite this paper: Yang Chen, Yaqian Wang, Da Chen, Jun Gu. (2025) Improvement of a Human Pose Estimation Strategy. Advances in Computer and Communication, 6(1), 41-47.
DOI: http://dx.doi.org/10.26855/acc.2025.01.007