Predicting 3D ground reaction force from 2D video via neural networks in sidestepping tasks

Corey Gene Morris; Marion Mundt; Molly Goldacre; Jason Weber; Ajmal Mian; Jacqueline Alderson

Predicting 3D ground reaction force from 2D video via neural networks in sidestepping tasks

Corey Gene Morris, Marion Mundt, Molly Goldacre, Jason Weber, Ajmal Mian, Jacqueline Alderson

Research output: Chapter in Book/Conference paper › Conference paper › peer-review

Abstract

Sports science practitioners often measure ground reaction forces (GRFs) to assess performance, rehabilitation and injury risk. However, recording of GRFs during dynamic tasks has historically been limited to lab settings. This work aims to use neural networks (NN) to predict three-dimensional (3D) GRF via pose estimation keypoints as inputs, determined from 2D video data. Two different NN were trained on a dataset containing 1474 samples from 14 participants and their prediction accuracy compared with ground truth force data. Results for both NN showed correlation coefficients ranging from 0.936 to 0.954 and normalised root mean square errors from 11.05% to 13.11% for anterior- posterior and vertical GRFs, with poorer results found in the medio-lateral direction. This study demonstrates the feasibility and utility of predicting GRFs from 2D video footage.

Original language	English
Title of host publication	ISBS Proceedings Archive
Pages	300-303
Volume	39
Publication status	Published - 6 Sept 2021
Event	39th Conference of the International Society of Biomechanics in Sport - Virtual, Canberra, Australia Duration: 3 Sept 2021 → 6 Sept 2021

Conference

Conference	39th Conference of the International Society of Biomechanics in Sport
Country/Territory	Australia
City	Canberra
Period	3/09/21 → 6/09/21

Access to Document

Cite this

@inproceedings{79357218c55b44129deecfb7955b5467,

title = "Predicting 3D ground reaction force from 2D video via neural networks in sidestepping tasks",

abstract = "Sports science practitioners often measure ground reaction forces (GRFs) to assess performance, rehabilitation and injury risk. However, recording of GRFs during dynamic tasks has historically been limited to lab settings. This work aims to use neural networks (NN) to predict three-dimensional (3D) GRF via pose estimation keypoints as inputs, determined from 2D video data. Two different NN were trained on a dataset containing 1474 samples from 14 participants and their prediction accuracy compared with ground truth force data. Results for both NN showed correlation coefficients ranging from 0.936 to 0.954 and normalised root mean square errors from 11.05% to 13.11% for anterior- posterior and vertical GRFs, with poorer results found in the medio-lateral direction. This study demonstrates the feasibility and utility of predicting GRFs from 2D video footage.",

author = "Morris, {Corey Gene} and Marion Mundt and Molly Goldacre and Jason Weber and Ajmal Mian and Jacqueline Alderson",

year = "2021",

month = sep,

day = "6",

language = "English",

volume = "39",

pages = "300--303",

booktitle = "ISBS Proceedings Archive",

note = "39th Conference of the International Society of Biomechanics in Sport ; Conference date: 03-09-2021 Through 06-09-2021",

}

Morris, CG, Mundt, M, Goldacre, M, Weber, J , Mian, A & Alderson, J 2021, Predicting 3D ground reaction force from 2D video via neural networks in sidestepping tasks. in ISBS Proceedings Archive. vol. 39, pp. 300-303, 39th Conference of the International Society of Biomechanics in Sport , Canberra, Australian Capital Territory, Australia, 3/09/21. <https://commons.nmu.edu/cgi/viewcontent.cgi?article=2215&context=isbs>

TY - GEN

T1 - Predicting 3D ground reaction force from 2D video via neural networks in sidestepping tasks

AU - Morris, Corey Gene

AU - Mundt, Marion

AU - Goldacre, Molly

AU - Weber, Jason

AU - Mian, Ajmal

AU - Alderson, Jacqueline

PY - 2021/9/6

Y1 - 2021/9/6

N2 - Sports science practitioners often measure ground reaction forces (GRFs) to assess performance, rehabilitation and injury risk. However, recording of GRFs during dynamic tasks has historically been limited to lab settings. This work aims to use neural networks (NN) to predict three-dimensional (3D) GRF via pose estimation keypoints as inputs, determined from 2D video data. Two different NN were trained on a dataset containing 1474 samples from 14 participants and their prediction accuracy compared with ground truth force data. Results for both NN showed correlation coefficients ranging from 0.936 to 0.954 and normalised root mean square errors from 11.05% to 13.11% for anterior- posterior and vertical GRFs, with poorer results found in the medio-lateral direction. This study demonstrates the feasibility and utility of predicting GRFs from 2D video footage.

AB - Sports science practitioners often measure ground reaction forces (GRFs) to assess performance, rehabilitation and injury risk. However, recording of GRFs during dynamic tasks has historically been limited to lab settings. This work aims to use neural networks (NN) to predict three-dimensional (3D) GRF via pose estimation keypoints as inputs, determined from 2D video data. Two different NN were trained on a dataset containing 1474 samples from 14 participants and their prediction accuracy compared with ground truth force data. Results for both NN showed correlation coefficients ranging from 0.936 to 0.954 and normalised root mean square errors from 11.05% to 13.11% for anterior- posterior and vertical GRFs, with poorer results found in the medio-lateral direction. This study demonstrates the feasibility and utility of predicting GRFs from 2D video footage.

M3 - Conference paper

VL - 39

SP - 300

EP - 303

BT - ISBS Proceedings Archive

T2 - 39th Conference of the International Society of Biomechanics in Sport

Y2 - 3 September 2021 through 6 September 2021

ER -

Predicting 3D ground reaction force from 2D video via neural networks in sidestepping tasks

Abstract

Conference

Access to Document

Fingerprint

Cite this