Human pose estimation in images and videos for sports analytics: 2D keypoint and 3D mesh estimation for challenging scenarios and extreme poses
- Since the emergence of professional sports competitions, athletes have continually sought to enhance their skills to achieve success. In recent decades, the widespread availability of cameras and smartphones has made video and image analysis an integral part of sports training, helping athletes to assess their performance and analyze potential for improvement. Since the rapid pace of athletic movements often makes it nearly impossible for the human eye to capture every detail in real time, even for coaches which excel at identifying errors and areas for improvement, video or image based analyses can help. Video recordings allow for detailed analysis through slow motion and repeated replays, enabling comprehensive feedback to athletes. However, such manual analyses are both time-intensive and laborious.
This thesis explores the application of computer vision techniques to automate motion analysis in the demanding context of sports. The focus of this thesis is on individual sports,Since the emergence of professional sports competitions, athletes have continually sought to enhance their skills to achieve success. In recent decades, the widespread availability of cameras and smartphones has made video and image analysis an integral part of sports training, helping athletes to assess their performance and analyze potential for improvement. Since the rapid pace of athletic movements often makes it nearly impossible for the human eye to capture every detail in real time, even for coaches which excel at identifying errors and areas for improvement, video or image based analyses can help. Video recordings allow for detailed analysis through slow motion and repeated replays, enabling comprehensive feedback to athletes. However, such manual analyses are both time-intensive and laborious.
This thesis explores the application of computer vision techniques to automate motion analysis in the demanding context of sports. The focus of this thesis is on individual sports, where tracking the athlete's posture is critical for movement evaluation. Specifically, we investigate the tasks of 2D Human Pose Estimation and 3D Human Mesh Estimation within the unique challenges posed by individual sports disciplines, including rapid motions and extreme body poses.…
Author: | Katja LudwigORCiDGND |
---|---|
URN: | urn:nbn:de:bvb:384-opus4-1226133 |
Frontdoor URL | https://opus.bibliothek.uni-augsburg.de/opus4/122613 |
Advisor: | Rainer Lienhart |
Type: | Doctoral Thesis |
Language: | English |
Year of first Publication: | 2025 |
Publishing Institution: | Universität Augsburg |
Granting Institution: | Universität Augsburg, Fakultät für Angewandte Informatik |
Date of final exam: | 2025/06/02 |
Release Date: | 2025/07/02 |
Tag: | Computer Vision; 2D Human Pose Estimation; 3D Human Mesh Estimation; Human Pose Estimation in Sports |
GND-Keyword: | Berufssport; Mensch; Pose; Maschinelles Sehen; Bildverarbeitung; Bewegungsanalyse |
Pagenumber: | 174 |
Institutes: | Fakultät für Angewandte Informatik |
Fakultät für Angewandte Informatik / Institut für Informatik | |
Fakultät für Angewandte Informatik / Institut für Informatik / Lehrstuhl für Maschinelles Lernen und Maschinelles Sehen | |
Dewey Decimal Classification: | 0 Informatik, Informationswissenschaft, allgemeine Werke / 00 Informatik, Wissen, Systeme / 004 Datenverarbeitung; Informatik |
Licence (German): | ![]() |