On the CPU Usage of Deep Learning Models on an Edge Device

Elarbi Badidi; Dhanya Gopinathan

doi:10.1007/978-3-031-21438-7_18

On the CPU Usage of Deep Learning Models on an Edge Device

Elarbi Badidi, Dhanya Gopinathan

Department of Computer Science and Software Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

2 Citations (Scopus)

Abstract

Applications that use the Internet of Things (IoT) capture massive amounts of raw data from sensors and actuators and frequently transmit this data to cloud data centers for processing and analysis. However, due to variable and unpredictable data generation rates and network latency, sending data to a cloud data center can result in a performance bottleneck. Data processing could occur at the network’s edge with the emergence of Fog and Edge computing-hosted microservices. Detecting and tracking objects from images, videos, and live streams are two of the fastest-growing computer vision applications increasingly being deployed at the edge. You Only Look Once (YOLO) models are highly optimized deep learning methods for object detection. This paper analyzes the CPU usage of four YOLO models on an edge device, an Nvidia Jetson Nano, at two different power budgets 5 and 10 W. Results show that the average CPU usage of the four YOLO models is low in 10 W power mode compared to 5 W power mode, except for YOLOv4-tiny. Furthermore, the number of frames per second processed by the four models remains relatively the same when switching from the 10 to 5 W power modes.

Original language	English
Title of host publication	Data Science and Algorithms in Systems - Proceedings of 6th Computational Methods in Systems and Software 2022, Vol. 2
Editors	Radek Silhavy, Petr Silhavy, Zdenka Prokopova
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	209-219
Number of pages	11
ISBN (Print)	9783031214370
DOIs	https://doi.org/10.1007/978-3-031-21438-7_18
Publication status	Published - 2023
Event	6th Computational Methods in Systems and Software, CoMeSySo 2022 - Virtual, Online Duration: Oct 10 2022 → Oct 15 2022

Publication series

Name	Lecture Notes in Networks and Systems
Volume	597 LNNS
ISSN (Print)	2367-3370
ISSN (Electronic)	2367-3389

Conference

Conference	6th Computational Methods in Systems and Software, CoMeSySo 2022
City	Virtual, Online
Period	10/10/22 → 10/15/22

Keywords

Deep learning
Edge computing
Edge intelligence
Intelligent traffic monitoring
Object detection

ASJC Scopus subject areas

Control and Systems Engineering
Signal Processing
Computer Networks and Communications

Access to Document

10.1007/978-3-031-21438-7_18

Cite this

Badidi, E., & Gopinathan, D. (2023). On the CPU Usage of Deep Learning Models on an Edge Device. In R. Silhavy, P. Silhavy, & Z. Prokopova (Eds.), Data Science and Algorithms in Systems - Proceedings of 6th Computational Methods in Systems and Software 2022, Vol. 2 (pp. 209-219). (Lecture Notes in Networks and Systems; Vol. 597 LNNS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-21438-7_18

On the CPU Usage of Deep Learning Models on an Edge Device. / Badidi, Elarbi; Gopinathan, Dhanya.
Data Science and Algorithms in Systems - Proceedings of 6th Computational Methods in Systems and Software 2022, Vol. 2. ed. / Radek Silhavy; Petr Silhavy; Zdenka Prokopova. Springer Science and Business Media Deutschland GmbH, 2023. p. 209-219 (Lecture Notes in Networks and Systems; Vol. 597 LNNS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Badidi, E & Gopinathan, D 2023, On the CPU Usage of Deep Learning Models on an Edge Device. in R Silhavy, P Silhavy & Z Prokopova (eds), Data Science and Algorithms in Systems - Proceedings of 6th Computational Methods in Systems and Software 2022, Vol. 2. Lecture Notes in Networks and Systems, vol. 597 LNNS, Springer Science and Business Media Deutschland GmbH, pp. 209-219, 6th Computational Methods in Systems and Software, CoMeSySo 2022, Virtual, Online, 10/10/22. https://doi.org/10.1007/978-3-031-21438-7_18

Badidi E, Gopinathan D. On the CPU Usage of Deep Learning Models on an Edge Device. In Silhavy R, Silhavy P, Prokopova Z, editors, Data Science and Algorithms in Systems - Proceedings of 6th Computational Methods in Systems and Software 2022, Vol. 2. Springer Science and Business Media Deutschland GmbH. 2023. p. 209-219. (Lecture Notes in Networks and Systems). doi: 10.1007/978-3-031-21438-7_18

Badidi, Elarbi ; Gopinathan, Dhanya. / On the CPU Usage of Deep Learning Models on an Edge Device. Data Science and Algorithms in Systems - Proceedings of 6th Computational Methods in Systems and Software 2022, Vol. 2. editor / Radek Silhavy ; Petr Silhavy ; Zdenka Prokopova. Springer Science and Business Media Deutschland GmbH, 2023. pp. 209-219 (Lecture Notes in Networks and Systems).

@inproceedings{f85f7a3d3aef4f69984926c9591c9e9e,

title = "On the CPU Usage of Deep Learning Models on an Edge Device",

abstract = "Applications that use the Internet of Things (IoT) capture massive amounts of raw data from sensors and actuators and frequently transmit this data to cloud data centers for processing and analysis. However, due to variable and unpredictable data generation rates and network latency, sending data to a cloud data center can result in a performance bottleneck. Data processing could occur at the network{\textquoteright}s edge with the emergence of Fog and Edge computing-hosted microservices. Detecting and tracking objects from images, videos, and live streams are two of the fastest-growing computer vision applications increasingly being deployed at the edge. You Only Look Once (YOLO) models are highly optimized deep learning methods for object detection. This paper analyzes the CPU usage of four YOLO models on an edge device, an Nvidia Jetson Nano, at two different power budgets 5 and 10 W. Results show that the average CPU usage of the four YOLO models is low in 10 W power mode compared to 5 W power mode, except for YOLOv4-tiny. Furthermore, the number of frames per second processed by the four models remains relatively the same when switching from the 10 to 5 W power modes.",

keywords = "Deep learning, Edge computing, Edge intelligence, Intelligent traffic monitoring, Object detection",

author = "Elarbi Badidi and Dhanya Gopinathan",

note = "Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 6th Computational Methods in Systems and Software, CoMeSySo 2022 ; Conference date: 10-10-2022 Through 15-10-2022",

year = "2023",

doi = "10.1007/978-3-031-21438-7_18",

language = "English",

isbn = "9783031214370",

series = "Lecture Notes in Networks and Systems",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "209--219",

editor = "Radek Silhavy and Petr Silhavy and Zdenka Prokopova",

booktitle = "Data Science and Algorithms in Systems - Proceedings of 6th Computational Methods in Systems and Software 2022, Vol. 2",

}

TY - GEN

T1 - On the CPU Usage of Deep Learning Models on an Edge Device

AU - Badidi, Elarbi

AU - Gopinathan, Dhanya

PY - 2023

Y1 - 2023

N2 - Applications that use the Internet of Things (IoT) capture massive amounts of raw data from sensors and actuators and frequently transmit this data to cloud data centers for processing and analysis. However, due to variable and unpredictable data generation rates and network latency, sending data to a cloud data center can result in a performance bottleneck. Data processing could occur at the network’s edge with the emergence of Fog and Edge computing-hosted microservices. Detecting and tracking objects from images, videos, and live streams are two of the fastest-growing computer vision applications increasingly being deployed at the edge. You Only Look Once (YOLO) models are highly optimized deep learning methods for object detection. This paper analyzes the CPU usage of four YOLO models on an edge device, an Nvidia Jetson Nano, at two different power budgets 5 and 10 W. Results show that the average CPU usage of the four YOLO models is low in 10 W power mode compared to 5 W power mode, except for YOLOv4-tiny. Furthermore, the number of frames per second processed by the four models remains relatively the same when switching from the 10 to 5 W power modes.

AB - Applications that use the Internet of Things (IoT) capture massive amounts of raw data from sensors and actuators and frequently transmit this data to cloud data centers for processing and analysis. However, due to variable and unpredictable data generation rates and network latency, sending data to a cloud data center can result in a performance bottleneck. Data processing could occur at the network’s edge with the emergence of Fog and Edge computing-hosted microservices. Detecting and tracking objects from images, videos, and live streams are two of the fastest-growing computer vision applications increasingly being deployed at the edge. You Only Look Once (YOLO) models are highly optimized deep learning methods for object detection. This paper analyzes the CPU usage of four YOLO models on an edge device, an Nvidia Jetson Nano, at two different power budgets 5 and 10 W. Results show that the average CPU usage of the four YOLO models is low in 10 W power mode compared to 5 W power mode, except for YOLOv4-tiny. Furthermore, the number of frames per second processed by the four models remains relatively the same when switching from the 10 to 5 W power modes.

KW - Deep learning

KW - Edge computing

KW - Edge intelligence

KW - Intelligent traffic monitoring

KW - Object detection

UR - http://www.scopus.com/inward/record.url?scp=85148749090&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85148749090&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-21438-7_18

DO - 10.1007/978-3-031-21438-7_18

M3 - Conference contribution

AN - SCOPUS:85148749090

SN - 9783031214370

T3 - Lecture Notes in Networks and Systems

SP - 209

EP - 219

BT - Data Science and Algorithms in Systems - Proceedings of 6th Computational Methods in Systems and Software 2022, Vol. 2

A2 - Silhavy, Radek

A2 - Silhavy, Petr

A2 - Prokopova, Zdenka

PB - Springer Science and Business Media Deutschland GmbH

T2 - 6th Computational Methods in Systems and Software, CoMeSySo 2022

Y2 - 10 October 2022 through 15 October 2022

ER -