A Proposed Method for Deep Learning-Based Automatic Tracking with Minimal Training Data for Sports Biomechanics Research

<b>Background:</b> This technical note proposes a deep learning-based, few-shot automatic key point tracking technique tailored to sports biomechanics research. <b>Methods:</b> The present method facilitates the arbitrary definition of key points on athletes’ bodies or sports...

Full description

Saved in:
Bibliographic Details
Main Authors: Daichi Yamashita, Minoru Matsumoto, Takeo Matsubayashi
Format: Article
Language:English
Published: MDPI AG 2025-04-01
Series:Biomechanics
Subjects:
Online Access:https://www.mdpi.com/2673-7078/5/2/25
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849433013485568000
author Daichi Yamashita
Minoru Matsumoto
Takeo Matsubayashi
author_facet Daichi Yamashita
Minoru Matsumoto
Takeo Matsubayashi
author_sort Daichi Yamashita
collection DOAJ
description <b>Background:</b> This technical note proposes a deep learning-based, few-shot automatic key point tracking technique tailored to sports biomechanics research. <b>Methods:</b> The present method facilitates the arbitrary definition of key points on athletes’ bodies or sports equipment. Initially, a limited number of video frames are manually digitized to mark the points of interest. These annotated frames are subsequently used to train a deep learning model that leverages a pre-trained VGG16 network as its backbone and incorporates an additional convolutional head. Feature maps extracted from three intermediate layers of VGG16 are processed by the head network to generate a probability map, highlighting the most likely locations of the key points. Transfer learning is implemented by freezing the backbone weights and training only the head network. By restricting the training data generation to regions surrounding the manually annotated points and training specifically for each video, this approach minimizes training time while maintaining high precision. <b>Conclusions:</b> This technique substantially reduces the time and effort required compared to frame-by-frame manual digitization in various sports settings, and enables customized training tailored to specific analytical needs and video environments.
format Article
id doaj-art-6406c5bebd2a4f58a46283d3527d1801
institution Kabale University
issn 2673-7078
language English
publishDate 2025-04-01
publisher MDPI AG
record_format Article
series Biomechanics
spelling doaj-art-6406c5bebd2a4f58a46283d3527d18012025-08-20T03:27:13ZengMDPI AGBiomechanics2673-70782025-04-01522510.3390/biomechanics5020025A Proposed Method for Deep Learning-Based Automatic Tracking with Minimal Training Data for Sports Biomechanics ResearchDaichi Yamashita0Minoru Matsumoto1Takeo Matsubayashi2Japan Institute of Sports Sciences, 3-15-1, Nishigaoka, Kita-ku, Tokyo 115-0056, JapanJapan Institute of Sports Sciences, 3-15-1, Nishigaoka, Kita-ku, Tokyo 115-0056, JapanJapan Institute of Sports Sciences, 3-15-1, Nishigaoka, Kita-ku, Tokyo 115-0056, Japan<b>Background:</b> This technical note proposes a deep learning-based, few-shot automatic key point tracking technique tailored to sports biomechanics research. <b>Methods:</b> The present method facilitates the arbitrary definition of key points on athletes’ bodies or sports equipment. Initially, a limited number of video frames are manually digitized to mark the points of interest. These annotated frames are subsequently used to train a deep learning model that leverages a pre-trained VGG16 network as its backbone and incorporates an additional convolutional head. Feature maps extracted from three intermediate layers of VGG16 are processed by the head network to generate a probability map, highlighting the most likely locations of the key points. Transfer learning is implemented by freezing the backbone weights and training only the head network. By restricting the training data generation to regions surrounding the manually annotated points and training specifically for each video, this approach minimizes training time while maintaining high precision. <b>Conclusions:</b> This technique substantially reduces the time and effort required compared to frame-by-frame manual digitization in various sports settings, and enables customized training tailored to specific analytical needs and video environments.https://www.mdpi.com/2673-7078/5/2/25motion analysisconvolutional neural networktransfer learningkey point detectionfew-shot learning
spellingShingle Daichi Yamashita
Minoru Matsumoto
Takeo Matsubayashi
A Proposed Method for Deep Learning-Based Automatic Tracking with Minimal Training Data for Sports Biomechanics Research
Biomechanics
motion analysis
convolutional neural network
transfer learning
key point detection
few-shot learning
title A Proposed Method for Deep Learning-Based Automatic Tracking with Minimal Training Data for Sports Biomechanics Research
title_full A Proposed Method for Deep Learning-Based Automatic Tracking with Minimal Training Data for Sports Biomechanics Research
title_fullStr A Proposed Method for Deep Learning-Based Automatic Tracking with Minimal Training Data for Sports Biomechanics Research
title_full_unstemmed A Proposed Method for Deep Learning-Based Automatic Tracking with Minimal Training Data for Sports Biomechanics Research
title_short A Proposed Method for Deep Learning-Based Automatic Tracking with Minimal Training Data for Sports Biomechanics Research
title_sort proposed method for deep learning based automatic tracking with minimal training data for sports biomechanics research
topic motion analysis
convolutional neural network
transfer learning
key point detection
few-shot learning
url https://www.mdpi.com/2673-7078/5/2/25
work_keys_str_mv AT daichiyamashita aproposedmethodfordeeplearningbasedautomatictrackingwithminimaltrainingdataforsportsbiomechanicsresearch
AT minorumatsumoto aproposedmethodfordeeplearningbasedautomatictrackingwithminimaltrainingdataforsportsbiomechanicsresearch
AT takeomatsubayashi aproposedmethodfordeeplearningbasedautomatictrackingwithminimaltrainingdataforsportsbiomechanicsresearch
AT daichiyamashita proposedmethodfordeeplearningbasedautomatictrackingwithminimaltrainingdataforsportsbiomechanicsresearch
AT minorumatsumoto proposedmethodfordeeplearningbasedautomatictrackingwithminimaltrainingdataforsportsbiomechanicsresearch
AT takeomatsubayashi proposedmethodfordeeplearningbasedautomatictrackingwithminimaltrainingdataforsportsbiomechanicsresearch