Multi-camera Torso Pose Estimation using Graph Neural Networks

Daniel Rodriguez-Criado; Pilar Bachiller; Pablo Bustos; George Vogiatzis; Luis J. Manso

doi:10.1109/RO-MAN47096.2020.9223542

Multi-camera Torso Pose Estimation using Graph Neural Networks

Daniel Rodriguez-Criado, Pilar Bachiller, Pablo Bustos, George Vogiatzis, Luis J. Manso^*

^*Corresponding author for this work

Research output: Chapter in Book/Published conference output › Conference publication

Abstract

Estimating the location and orientation of humans is an essential skill for service and assistive robots. To achieve a reliable estimation in a wide area such as an apartment, multiple RGBD cameras are frequently used. Firstly, these setups are relatively expensive. Secondly, they seldom perform an effective data fusion using the multiple camera sources at an early stage of the processing pipeline. Occlusions and partial views make this second point very relevant in these scenarios. The proposal presented in this paper makes use of graph neural networks to merge the information acquired from multiple camera sources, achieving a mean absolute error below 125 mm for the location and 10 degrees for the orientation using low-resolution RGB images. The experiments, conducted in an apartment with three cameras, benchmarked two different graph neural network implementations and a third architecture based on fully connected layers. The software used has been released as open-source in a public repository.

Original language	English
Title of host publication	29th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2020
Publisher	IEEE
Pages	827-832
Number of pages	6
ISBN (Electronic)	978-1-7281-6075-7
ISBN (Print)	978-1-7281-6076-4
DOIs	https://doi.org/10.1109/RO-MAN47096.2020.9223542
Publication status	Published - 14 Oct 2020
Event	IEEE International Conference on Robot & Human Interactive Communication - Virtual, Italy Duration: 31 Aug 2020 → 4 Sept 2020 Conference number: 29th http://ro-man2020.unina.it

Publication series

Name	29th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2020

Conference

Conference	IEEE International Conference on Robot & Human Interactive Communication
Abbreviated title	RO-MAN
Country/Territory	Italy
City	Virtual
Period	31/08/20 → 4/09/20
Internet address	http://ro-man2020.unina.it

Bibliographical note

CC BY-SA
© 2020 The Authors

Keywords

human tracking
graph neural networks
sensorised environments

Access to Document

10.1109/RO-MAN47096.2020.9223542

Multi-camera Torso Pose Estimation using Graph Neural Networks
CC BY-SA © 2020 The Authors
Accepted author manuscript, 665 KBLicence: CC BY-SA 3.0

Cite this

Rodriguez-Criado, D., Bachiller, P., Bustos, P., Vogiatzis, G., & Manso, L. J. (2020). Multi-camera Torso Pose Estimation using Graph Neural Networks. In 29th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2020 (pp. 827-832). Article 9223542 (29th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2020). IEEE. https://doi.org/10.1109/RO-MAN47096.2020.9223542

@inproceedings{92001db61591439d922438da87f6ed7a,

title = "Multi-camera Torso Pose Estimation using Graph Neural Networks",

abstract = "Estimating the location and orientation of humans is an essential skill for service and assistive robots. To achieve a reliable estimation in a wide area such as an apartment, multiple RGBD cameras are frequently used. Firstly, these setups are relatively expensive. Secondly, they seldom perform an effective data fusion using the multiple camera sources at an early stage of the processing pipeline. Occlusions and partial views make this second point very relevant in these scenarios. The proposal presented in this paper makes use of graph neural networks to merge the information acquired from multiple camera sources, achieving a mean absolute error below 125 mm for the location and 10 degrees for the orientation using low-resolution RGB images. The experiments, conducted in an apartment with three cameras, benchmarked two different graph neural network implementations and a third architecture based on fully connected layers. The software used has been released as open-source in a public repository. ",

keywords = "human tracking, graph neural networks, sensorised environments",

author = "Daniel Rodriguez-Criado and Pilar Bachiller and Pablo Bustos and George Vogiatzis and Manso, {Luis J.}",

note = "CC BY-SA {\textcopyright} 2020 The Authors; IEEE International Conference on Robot & Human Interactive Communication, RO-MAN ; Conference date: 31-08-2020 Through 04-09-2020",

year = "2020",

month = oct,

day = "14",

doi = "10.1109/RO-MAN47096.2020.9223542",

language = "English",

isbn = "978-1-7281-6076-4",

series = "29th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2020",

publisher = "IEEE",

pages = "827--832",

booktitle = "29th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2020",

address = "United States",

url = "http://ro-man2020.unina.it",

}

Rodriguez-Criado, D, Bachiller, P, Bustos, P, Vogiatzis, G & Manso, LJ 2020, Multi-camera Torso Pose Estimation using Graph Neural Networks. in 29th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2020., 9223542, 29th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2020, IEEE, pp. 827-832, IEEE International Conference on Robot & Human Interactive Communication, Virtual, Italy, 31/08/20. https://doi.org/10.1109/RO-MAN47096.2020.9223542

Multi-camera Torso Pose Estimation using Graph Neural Networks. / Rodriguez-Criado, Daniel; Bachiller, Pilar; Bustos, Pablo et al.
29th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2020. IEEE, 2020. p. 827-832 9223542 (29th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2020).

Research output: Chapter in Book/Published conference output › Conference publication

TY - GEN

T1 - Multi-camera Torso Pose Estimation using Graph Neural Networks

AU - Rodriguez-Criado, Daniel

AU - Bachiller, Pilar

AU - Bustos, Pablo

AU - Vogiatzis, George

AU - Manso, Luis J.

N1 - Conference code: 29th

PY - 2020/10/14

Y1 - 2020/10/14

N2 - Estimating the location and orientation of humans is an essential skill for service and assistive robots. To achieve a reliable estimation in a wide area such as an apartment, multiple RGBD cameras are frequently used. Firstly, these setups are relatively expensive. Secondly, they seldom perform an effective data fusion using the multiple camera sources at an early stage of the processing pipeline. Occlusions and partial views make this second point very relevant in these scenarios. The proposal presented in this paper makes use of graph neural networks to merge the information acquired from multiple camera sources, achieving a mean absolute error below 125 mm for the location and 10 degrees for the orientation using low-resolution RGB images. The experiments, conducted in an apartment with three cameras, benchmarked two different graph neural network implementations and a third architecture based on fully connected layers. The software used has been released as open-source in a public repository.

AB - Estimating the location and orientation of humans is an essential skill for service and assistive robots. To achieve a reliable estimation in a wide area such as an apartment, multiple RGBD cameras are frequently used. Firstly, these setups are relatively expensive. Secondly, they seldom perform an effective data fusion using the multiple camera sources at an early stage of the processing pipeline. Occlusions and partial views make this second point very relevant in these scenarios. The proposal presented in this paper makes use of graph neural networks to merge the information acquired from multiple camera sources, achieving a mean absolute error below 125 mm for the location and 10 degrees for the orientation using low-resolution RGB images. The experiments, conducted in an apartment with three cameras, benchmarked two different graph neural network implementations and a third architecture based on fully connected layers. The software used has been released as open-source in a public repository.

KW - human tracking

KW - graph neural networks

KW - sensorised environments

UR - https://arxiv.org/abs/2007.14126

UR - http://www.scopus.com/inward/record.url?scp=85095766187&partnerID=8YFLogxK

U2 - 10.1109/RO-MAN47096.2020.9223542

DO - 10.1109/RO-MAN47096.2020.9223542

M3 - Conference publication

SN - 978-1-7281-6076-4

T3 - 29th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2020

SP - 827

EP - 832

BT - 29th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2020

PB - IEEE

T2 - IEEE International Conference on Robot & Human Interactive Communication

Y2 - 31 August 2020 through 4 September 2020

ER -

Rodriguez-Criado D, Bachiller P, Bustos P, Vogiatzis G, Manso LJ. Multi-camera Torso Pose Estimation using Graph Neural Networks. In 29th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2020. IEEE. 2020. p. 827-832. 9223542. (29th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2020). doi: 10.1109/RO-MAN47096.2020.9223542

Multi-camera Torso Pose Estimation using Graph Neural Networks

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this