Research Article
Challenges in building machine learning models for movement recognition in sports caused by technique inconsistencies of beginners
@INPROCEEDINGS{10.4108/eai.18-12-2025.2365256, author={Val Vec and Saso Tomažič and Anton Kos and Anton Umek}, title={Challenges in building machine learning models for movement recognition in sports caused by technique inconsistencies of beginners}, proceedings={Proceedings of the 13th International Conference on Identification, Information and Knowledge in the Internet of Things, IIKI 2025, 18-21 December 2025, Chengdu, China}, publisher={EAI}, proceedings_a={IIKI}, year={2026}, month={6}, keywords={technique inconsistencies biomechanical feedback machine learning person recognition}, doi={10.4108/eai.18-12-2025.2365256} }- Val Vec
Saso Tomažič
Anton Kos
Anton Umek
Year: 2026
Challenges in building machine learning models for movement recognition in sports caused by technique inconsistencies of beginners
IIKI
EAI
DOI: 10.4108/eai.18-12-2025.2365256
Abstract
This research is part of a broader effort to develop machine learning–based systems that provide real-time feedback to athletes during training. In this paper, we focus on inconsistencies in the movement techniques of beginner athletes, using dart throwing as a case study. In the first part of the paper, we demonstrate that a beginner’s technique changes from day to day. We show that these changes are gradual rather than random, suggesting that both learning and forgetting play a role. In the subsequent sections, we investigate how this variability affects two tasks: identifying the individual and detecting poor technique using machine learning methods. Our findings suggest that the variability in beginners’ techniques is substantial enough that methods designed for professional athletes are not directly applicable when building systems for novices.
