Towards Machine Learning Framework for Badminton Game Analysis Using TrackNet and YOLO Models

Thumbnail Image
Date
2023-05
Authors
Mohamed, Amna
Major Professor
Mitra, Simanta
Prabhu, Gurpur M
Advisor
Committee Member
Mitra, Simanta
Prabhu, Gurpur M
Journal Title
Journal ISSN
Volume Title
Publisher
Authors
Research Projects
Organizational Units
Journal Issue
Is Version Of
Versions
Series
Department
Abstract
Badminton is a popular sport played worldwide, and analyzing the game can provide valuable insights for players, coaches, and researchers. The current methods available for analyzing badminton games can be costly and resource-intensive. Machine learning methods have the potential to automate and enhance the analysis of badminton games. The low-cost implementation of these methods makes them accessible to a wider audience, providing an affordable option for analysis and improvement. In this report, we explore the use of popular object-tracking machine learning methods - TrackNet and YOLO - to analyze badminton games from a single smartphone-recorded video of the game. TrackNet is a deep learning-based algorithm that can track the trajectory of a shuttlecock, while YOLO is an object detection algorithm that can identify players and their positions on the court. Using a custom dataset of badminton games, we evaluate the performance of these methods in terms of accuracy and computational efficiency in tracking the shuttlecock, tracking players, and identifying shot types. Our results show that TrackNet can accurately predict the trajectory of the shuttlecock, while the YOLO model can identify players and detect shot types with high precision. This research introduces a cost-effective analysis framework that represents the initial stage in developing a recommendation system aimed at amateur players to enhance their gameplay.
Comments
Description
Keywords
Citation
DOI
Source
Subject Categories
Copyright
2023