Research on Human Pose Estimation Based on 2D and 3D Classification

Xu Ji, Can Liu, Lingyue Zeng

2024

Abstract

Human pose estimation (HPE) refers to the use of computer vision technology and machine learning methods to detect and analyse human pose information from images or videos. The early mainly focused on solving single person pose estimation problems. It treats the problems as classification or regression problems using a global feature. However, such methods are not accurate enough and only suitable for simple scenarios. The application of HPE estimation based on deep learning is very promising and can provide many important advantages and innovations in related field. Therefore, this paper mainly focuses on several methods and technical advances for 2D and 3D HPE based on deep learning. This paper aims to introduce the technique and compare the characteristics of different categories. Besides, it covers various datasets since 2015 and different metrics for evaluating methods performance and illustrating considerable potential for future development. Finally, this paper discusses the advantages and limitations through comparison of methods. With the development of computer processing ability, the future human pose estimation tasks will make more progress in improving accuracy, expanding application scenarios and optimizing efficiency.

Paper Citation

in Harvard Style

Ji X., Liu C. and Zeng L. (2024). Research on Human Pose Estimation Based on 2D and 3D Classification. In Proceedings of the 1st International Conference on Modern Logistics and Supply Chain Management - Volume 1: MLSCM; ISBN 978-989-758-738-2, SciTePress, pages 99-105. DOI: 10.5220/0013234200004558

in Bibtex Style

@conference{mlscm24,
author={Xu Ji and Can Liu and Lingyue Zeng},
title={Research on Human Pose Estimation Based on 2D and 3D Classification},
booktitle={Proceedings of the 1st International Conference on Modern Logistics and Supply Chain Management - Volume 1: MLSCM},
year={2024},
pages={99-105},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013234200004558},
isbn={978-989-758-738-2},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Modern Logistics and Supply Chain Management - Volume 1: MLSCM
TI - Research on Human Pose Estimation Based on 2D and 3D Classification
SN - 978-989-758-738-2
AU - Ji X.
AU - Liu C.
AU - Zeng L.
PY - 2024
SP - 99
EP - 105
DO - 10.5220/0013234200004558
PB - SciTePress