YOLOv3-mobile for real-time pedestrian detection on embedded GPU

Mohammad Alkhaleefah; Narendra Babu Tatini; Hung Te Lee; Tan Hsu Tan; Shang Chih Ma; Yang Lang Chang

doi:10.1145/3474906.3474915

YOLOv3-mobile for real-time pedestrian detection on embedded GPU

Mohammad Alkhaleefah
, Narendra Babu Tatini
, Hung Te Lee
, Tan Hsu Tan
, Shang Chih Ma
, Yang Lang Chang

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

3 Scopus citations

Abstract

Pedestrian detection is one of the challenging tasks in the technology of autonomous driving. Recently, the object detection network of you only look once (YOLO), especially YOLOv3 and YOLOv3-tiny have demonstrated a high level of pedestrian detection performance on a powerful GPU card such as Pascal Titan X. However, it is still challenging to use YOLOv3 and YOLOv3-tiny on embedded GPU system due to their large network size. In this paper, we present a lightweight YOLOv3-mobile network by refining the architecture of YOLOv3-tiny to improve its pedestrian detection efficiency on embedded GPUs such as Nvidia Jetson TX1. The experimental results showed that the proposed framework can accelerate the frame rate per second (FPS) from 18 FPS to 37 FPS with comparable mean average precision (mAP).

Original language	English
Title of host publication	ICGSP 2021 - 5th International Conference on Graphics and Signal Processing
Publisher	Association for Computing Machinery
Pages	27-31
Number of pages	5
ISBN (Electronic)	9781450389419
DOIs	https://doi.org/10.1145/3474906.3474915
State	Published - 25 Jun 2021
Event	5th International Conference on Graphics and Signal Processing, ICGSP 2021 - Virtual, Online, Japan Duration: 25 Jun 2021 → 27 Jun 2021

Publication series

Name	ACM International Conference Proceeding Series

Conference

Conference	5th International Conference on Graphics and Signal Processing, ICGSP 2021
Country/Territory	Japan
City	Virtual, Online
Period	25/06/21 → 27/06/21

Bibliographical note

Publisher Copyright:
© 2021 ACM.

Keywords

CUDA
Embedded GPU
Real-time pedestrian detection
YOLOv3-mobile

Access to Document

10.1145/3474906.3474915

Cite this

Alkhaleefah, M., Tatini, N. B., Lee, H. T., Tan, T. H., Ma, S. C., & Chang, Y. L. (2021). YOLOv3-mobile for real-time pedestrian detection on embedded GPU. In ICGSP 2021 - 5th International Conference on Graphics and Signal Processing (pp. 27-31). (ACM International Conference Proceeding Series). Association for Computing Machinery. https://doi.org/10.1145/3474906.3474915

@inproceedings{e8265d2be6714cfb8c9630292c29f090,

title = "YOLOv3-mobile for real-time pedestrian detection on embedded GPU",

abstract = "Pedestrian detection is one of the challenging tasks in the technology of autonomous driving. Recently, the object detection network of you only look once (YOLO), especially YOLOv3 and YOLOv3-tiny have demonstrated a high level of pedestrian detection performance on a powerful GPU card such as Pascal Titan X. However, it is still challenging to use YOLOv3 and YOLOv3-tiny on embedded GPU system due to their large network size. In this paper, we present a lightweight YOLOv3-mobile network by refining the architecture of YOLOv3-tiny to improve its pedestrian detection efficiency on embedded GPUs such as Nvidia Jetson TX1. The experimental results showed that the proposed framework can accelerate the frame rate per second (FPS) from 18 FPS to 37 FPS with comparable mean average precision (mAP).",

keywords = "CUDA, Embedded GPU, Real-time pedestrian detection, YOLOv3-mobile",

author = "Mohammad Alkhaleefah and Tatini, \{Narendra Babu\} and Lee, \{Hung Te\} and Tan, \{Tan Hsu\} and Ma, \{Shang Chih\} and Chang, \{Yang Lang\}",

note = "Publisher Copyright: {\textcopyright} 2021 ACM.; 5th International Conference on Graphics and Signal Processing, ICGSP 2021 ; Conference date: 25-06-2021 Through 27-06-2021",

year = "2021",

month = jun,

day = "25",

doi = "10.1145/3474906.3474915",

language = "???core.languages.en\_GB???",

series = "ACM International Conference Proceeding Series",

publisher = "Association for Computing Machinery",

pages = "27--31",

booktitle = "ICGSP 2021 - 5th International Conference on Graphics and Signal Processing",

}

Alkhaleefah, M, Tatini, NB, Lee, HT, Tan, TH , Ma, SC & Chang, YL 2021, YOLOv3-mobile for real-time pedestrian detection on embedded GPU. in ICGSP 2021 - 5th International Conference on Graphics and Signal Processing. ACM International Conference Proceeding Series, Association for Computing Machinery, pp. 27-31, 5th International Conference on Graphics and Signal Processing, ICGSP 2021, Virtual, Online, Japan, 25/06/21. https://doi.org/10.1145/3474906.3474915

YOLOv3-mobile for real-time pedestrian detection on embedded GPU. / Alkhaleefah, Mohammad; Tatini, Narendra Babu; Lee, Hung Te et al.
ICGSP 2021 - 5th International Conference on Graphics and Signal Processing. Association for Computing Machinery, 2021. p. 27-31 (ACM International Conference Proceeding Series).

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

TY - GEN

T1 - YOLOv3-mobile for real-time pedestrian detection on embedded GPU

AU - Alkhaleefah, Mohammad

AU - Tatini, Narendra Babu

AU - Lee, Hung Te

AU - Tan, Tan Hsu

AU - Ma, Shang Chih

AU - Chang, Yang Lang

PY - 2021/6/25

Y1 - 2021/6/25

N2 - Pedestrian detection is one of the challenging tasks in the technology of autonomous driving. Recently, the object detection network of you only look once (YOLO), especially YOLOv3 and YOLOv3-tiny have demonstrated a high level of pedestrian detection performance on a powerful GPU card such as Pascal Titan X. However, it is still challenging to use YOLOv3 and YOLOv3-tiny on embedded GPU system due to their large network size. In this paper, we present a lightweight YOLOv3-mobile network by refining the architecture of YOLOv3-tiny to improve its pedestrian detection efficiency on embedded GPUs such as Nvidia Jetson TX1. The experimental results showed that the proposed framework can accelerate the frame rate per second (FPS) from 18 FPS to 37 FPS with comparable mean average precision (mAP).

AB - Pedestrian detection is one of the challenging tasks in the technology of autonomous driving. Recently, the object detection network of you only look once (YOLO), especially YOLOv3 and YOLOv3-tiny have demonstrated a high level of pedestrian detection performance on a powerful GPU card such as Pascal Titan X. However, it is still challenging to use YOLOv3 and YOLOv3-tiny on embedded GPU system due to their large network size. In this paper, we present a lightweight YOLOv3-mobile network by refining the architecture of YOLOv3-tiny to improve its pedestrian detection efficiency on embedded GPUs such as Nvidia Jetson TX1. The experimental results showed that the proposed framework can accelerate the frame rate per second (FPS) from 18 FPS to 37 FPS with comparable mean average precision (mAP).

KW - CUDA

KW - Embedded GPU

KW - Real-time pedestrian detection

KW - YOLOv3-mobile

UR - https://www.scopus.com/pages/publications/85118220724

U2 - 10.1145/3474906.3474915

DO - 10.1145/3474906.3474915

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:85118220724

T3 - ACM International Conference Proceeding Series

SP - 27

EP - 31

BT - ICGSP 2021 - 5th International Conference on Graphics and Signal Processing

PB - Association for Computing Machinery

T2 - 5th International Conference on Graphics and Signal Processing, ICGSP 2021

Y2 - 25 June 2021 through 27 June 2021

ER -

YOLOv3-mobile for real-time pedestrian detection on embedded GPU

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this