Front Moving Object Behavior Prediction System Exploiting Deep Learning Technology for ADAS Applications

Wen Chia Tsai; Kuan Chou Chen; Jhih Sheng Lai; Jiun In Guo

doi:10.1109/MWSCAS48704.2020.9184578

Front Moving Object Behavior Prediction System Exploiting Deep Learning Technology for ADAS Applications

Wen Chia Tsai, Kuan Chou Chen, Jhih Sheng Lai, Jiun In Guo^*

^*Corresponding author for this work

Institute of Electronics

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

1 Scopus citations

Abstract

This paper proposes a front pedestrian crossing and vehicle cut-in prediction system based on 3D convolution behavior prediction network. The proposed design improves the original 3D convolution network (C3D) to make behavior recognition network have the ability of object localization, which is important to detect multiple moving object behaviors. The proposed system is implemented on the embedded system in real-time, which achieves 20 frames per second when it is deployed on NVIDIA Jetson AGX Xavier and possesses over 92.8% accuracy for pedestrian crossing and 94.3% accuracy for vehicle cut-in behavior detection.

Original language	English
Title of host publication	2020 IEEE 63rd International Midwest Symposium on Circuits and Systems, MWSCAS 2020 - Proceedings
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	1052-1055
Number of pages	4
ISBN (Electronic)	9781538629161
DOIs	https://doi.org/10.1109/MWSCAS48704.2020.9184578
State	Published - Aug 2020
Event	63rd IEEE International Midwest Symposium on Circuits and Systems, MWSCAS 2020 - Springfield, United States Duration: 9 Aug 2020 → 12 Aug 2020

Publication series

Name	Midwest Symposium on Circuits and Systems
Volume	2020-August
ISSN (Print)	1548-3746

Conference

Conference	63rd IEEE International Midwest Symposium on Circuits and Systems, MWSCAS 2020
Country/Territory	United States
City	Springfield
Period	9/08/20 → 12/08/20

Keywords

behavior recognition
deep learning
Pedestrian detection
vehicle cut-in

Access to Document

10.1109/MWSCAS48704.2020.9184578

Cite this

Tsai, W. C., Chen, K. C., Lai, J. S., & Guo, J. I. (2020). Front Moving Object Behavior Prediction System Exploiting Deep Learning Technology for ADAS Applications. In 2020 IEEE 63rd International Midwest Symposium on Circuits and Systems, MWSCAS 2020 - Proceedings (pp. 1052-1055). Article 9184578 (Midwest Symposium on Circuits and Systems; Vol. 2020-August). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/MWSCAS48704.2020.9184578

Tsai, Wen Chia ; Chen, Kuan Chou ; Lai, Jhih Sheng et al. / Front Moving Object Behavior Prediction System Exploiting Deep Learning Technology for ADAS Applications. 2020 IEEE 63rd International Midwest Symposium on Circuits and Systems, MWSCAS 2020 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2020. pp. 1052-1055 (Midwest Symposium on Circuits and Systems).

@inproceedings{4d4595813c014b549bc1dc2ac1243c49,

title = "Front Moving Object Behavior Prediction System Exploiting Deep Learning Technology for ADAS Applications",

abstract = "This paper proposes a front pedestrian crossing and vehicle cut-in prediction system based on 3D convolution behavior prediction network. The proposed design improves the original 3D convolution network (C3D) to make behavior recognition network have the ability of object localization, which is important to detect multiple moving object behaviors. The proposed system is implemented on the embedded system in real-time, which achieves 20 frames per second when it is deployed on NVIDIA Jetson AGX Xavier and possesses over 92.8% accuracy for pedestrian crossing and 94.3% accuracy for vehicle cut-in behavior detection.",

keywords = "behavior recognition, deep learning, Pedestrian detection, vehicle cut-in",

author = "Tsai, {Wen Chia} and Chen, {Kuan Chou} and Lai, {Jhih Sheng} and Guo, {Jiun In}",

year = "2020",

month = aug,

doi = "10.1109/MWSCAS48704.2020.9184578",

language = "English",

series = "Midwest Symposium on Circuits and Systems",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "1052--1055",

booktitle = "2020 IEEE 63rd International Midwest Symposium on Circuits and Systems, MWSCAS 2020 - Proceedings",

address = "United States",

note = "63rd IEEE International Midwest Symposium on Circuits and Systems, MWSCAS 2020 ; Conference date: 09-08-2020 Through 12-08-2020",

}

Tsai, WC, Chen, KC, Lai, JS & Guo, JI 2020, Front Moving Object Behavior Prediction System Exploiting Deep Learning Technology for ADAS Applications. in 2020 IEEE 63rd International Midwest Symposium on Circuits and Systems, MWSCAS 2020 - Proceedings., 9184578, Midwest Symposium on Circuits and Systems, vol. 2020-August, Institute of Electrical and Electronics Engineers Inc., pp. 1052-1055, 63rd IEEE International Midwest Symposium on Circuits and Systems, MWSCAS 2020, Springfield, United States, 9/08/20. https://doi.org/10.1109/MWSCAS48704.2020.9184578

Front Moving Object Behavior Prediction System Exploiting Deep Learning Technology for ADAS Applications. / Tsai, Wen Chia; Chen, Kuan Chou; Lai, Jhih Sheng et al.
2020 IEEE 63rd International Midwest Symposium on Circuits and Systems, MWSCAS 2020 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2020. p. 1052-1055 9184578 (Midwest Symposium on Circuits and Systems; Vol. 2020-August).

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

TY - GEN

T1 - Front Moving Object Behavior Prediction System Exploiting Deep Learning Technology for ADAS Applications

AU - Tsai, Wen Chia

AU - Chen, Kuan Chou

AU - Lai, Jhih Sheng

AU - Guo, Jiun In

PY - 2020/8

Y1 - 2020/8

N2 - This paper proposes a front pedestrian crossing and vehicle cut-in prediction system based on 3D convolution behavior prediction network. The proposed design improves the original 3D convolution network (C3D) to make behavior recognition network have the ability of object localization, which is important to detect multiple moving object behaviors. The proposed system is implemented on the embedded system in real-time, which achieves 20 frames per second when it is deployed on NVIDIA Jetson AGX Xavier and possesses over 92.8% accuracy for pedestrian crossing and 94.3% accuracy for vehicle cut-in behavior detection.

AB - This paper proposes a front pedestrian crossing and vehicle cut-in prediction system based on 3D convolution behavior prediction network. The proposed design improves the original 3D convolution network (C3D) to make behavior recognition network have the ability of object localization, which is important to detect multiple moving object behaviors. The proposed system is implemented on the embedded system in real-time, which achieves 20 frames per second when it is deployed on NVIDIA Jetson AGX Xavier and possesses over 92.8% accuracy for pedestrian crossing and 94.3% accuracy for vehicle cut-in behavior detection.

KW - behavior recognition

KW - deep learning

KW - Pedestrian detection

KW - vehicle cut-in

UR - http://www.scopus.com/inward/record.url?scp=85090560428&partnerID=8YFLogxK

U2 - 10.1109/MWSCAS48704.2020.9184578

DO - 10.1109/MWSCAS48704.2020.9184578

M3 - Conference contribution

AN - SCOPUS:85090560428

T3 - Midwest Symposium on Circuits and Systems

SP - 1052

EP - 1055

BT - 2020 IEEE 63rd International Midwest Symposium on Circuits and Systems, MWSCAS 2020 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 63rd IEEE International Midwest Symposium on Circuits and Systems, MWSCAS 2020

Y2 - 9 August 2020 through 12 August 2020

ER -

Tsai WC, Chen KC, Lai JS, Guo JI. Front Moving Object Behavior Prediction System Exploiting Deep Learning Technology for ADAS Applications. In 2020 IEEE 63rd International Midwest Symposium on Circuits and Systems, MWSCAS 2020 - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2020. p. 1052-1055. 9184578. (Midwest Symposium on Circuits and Systems). doi: 10.1109/MWSCAS48704.2020.9184578

Front Moving Object Behavior Prediction System Exploiting Deep Learning Technology for ADAS Applications

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this