Related Experiment Video
Updated: Jan 7, 2026
05:41
A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis
Published on: February 6, 2020
9.8K
The Application of Deep Learning Human Pose Estimation in Sport: A Systematic Review
Cavan Aulton1, Lois Wakili2, Ben William Strafford2
1School of Sport and Physical Activity, College of Health, Wellbeing and Life Sciences, Sheffield Hallam University, Collegiate Hall, Collegiate Crescent, Sheffield, S10 2BP, UK. cavan_aulton@yahoo.co.uk.
Sports Medicine - Open
|December 10, 2025
Summary
Deep Learning-based Human Pose Estimation (HPE) offers insights into sports movement analysis. However, limited open datasets and reproducibility hinder widespread adoption, necessitating standardized practices for future advancements in sports science.
Area of Science:
- Sports Science
- Computer Vision
- Biomechanics
Background:
- Deep Learning (DL)-based Human Pose Estimation (HPE) is increasingly vital in sports research for precise joint localization in visual data.
- DL-based HPE enables non-invasive analysis of movement for training, performance optimization, and injury prevention.
Purpose of the Study:
- To systematically review the application of DL-based HPE in sports.
- To assess the availability and accessibility of training datasets, reproducibility, and human factors.
- To provide recommendations for future research and applications in sports science.
Main Methods:
- A systematic literature search following PRISMA guidelines was conducted across four major databases.
- 371 articles were screened, with two independent reviewers applying inclusion/exclusion criteria.
- Data were descriptively synthesized, focusing on DL-based HPE applications, dataset characteristics, and algorithms.
Main Results:
- DL-based HPE applications in sports include movement skill analysis, action recognition, coaching tools, and officiating support.
- Most studies utilized private datasets, limiting reproducibility and generalizability.
- Bespoke multi-model algorithms and single-person pose estimation were common, but challenges remain for practical implementation.
Conclusions:
- This review is the first systematic evaluation of DL-based HPE from a sports science perspective.
- The field requires open, standardized datasets and reproducible methodologies for advancement.
- Future research should address limitations and explore innovative applications to maximize DL-based HPE's impact in sports science.
