Multi-Person Pose Estimation: Method Classification and Cross-Dataset Performance Analysis

DOI

10.2991/978-94-6239-648-7_16How to use a DOI?

Keywords

Multi-person pose estimation; Deep learning; Top-down; Bottom-up; Vector field regression

Abstract

Finding the important features of each human body in the picture and accurately allocating those features to each individual is the main challenge of multi-person posture estimation. Multi person pose estimation tasks can provide support for multiple downstream tasks and overcome the limitation of single person pose estimation that can only recognize a single human body. As deep learning has advanced in the field of machine vision, deep learning-based multi-person pose estimation techniques have progressively supplanted conventional techniques and are now widely used. The task is divided into two methods based on human body modeling: heatmap and vector field estimation. Among them, the heatmap method is further divided into top-down and bottom-up methods based on modeling logic. This article first divides the task of multi-person pose estimation into heatmap and vector field regression, and analyzes their characteristics by category. It then compares their performance on two datasets. Finally, it offers a prospect for future development. This article summarizes the advantages, disadvantages and main improvement directions of different types of implementation methods. Its comparison study also identifies the approaches’ performance focus, which can serve as a guide for further research.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Zikun Li
PY  - 2026
DA  - 2026/04/24
TI  - Multi-Person Pose Estimation: Method Classification and Cross-Dataset Performance Analysis
BT  - Proceedings of the International Workshop on Advances in Deep Learning for Image Analysis and Computer Vision (IWADIC 2025)
PB  - Atlantis Press
SP  - 136
EP  - 144
SN  - 2352-538X
UR  - https://doi.org/10.2991/978-94-6239-648-7_16
DO  - 10.2991/978-94-6239-648-7_16
ID  - Li2026
ER  -