Efficient Light Deep Network for Street Scene Parsing

Zhe Hui Wang; Sanyuan Zhao; Jianbing Shen; Zhengchao Lei

doi:10.1109/VCIP49819.2020.9301795

Efficient Light Deep Network for Street Scene Parsing

Zhe Hui Wang, Sanyuan Zhao, Jianbing Shen, Zhengchao Lei

School of Computer Science and Technology

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

Abstract

The semantic segmentation is a dense pixel label pre-diction task, which takes quite a lot of resources and computation cost in most of the time. In our approach, we pay attention to balance the speed and better performance which outperforms the state of the art in speed and accuracy for real-time performance. We come up with the idea of new efficient deep backbone that can extract more semantic details, reduce the computation cost and be easy to deploy at the same time. We call our new backbone as Cascaded Mobile Network, which is proved to be very useful. Our proposed model achieves 72.1 mIOU on the CityScapes val, and 69.5 on CamVid. We achieve good balance between speed and accuracy.

Original language	English
Title of host publication	2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	42-45
Number of pages	4
ISBN (Electronic)	9781728180670
DOIs	https://doi.org/10.1109/VCIP49819.2020.9301795
Publication status	Published - 1 Dec 2020
Event	2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020 - Virtual, Macau, China Duration: 1 Dec 2020 → 4 Dec 2020

Publication series

Name	2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020

Conference

Conference	2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020
Country/Territory	China
City	Virtual, Macau
Period	1/12/20 → 4/12/20

Keywords

CNN
Deep Learning
Efficient B ackbone
Semantic Segmentation
real-time

Access to Document

10.1109/VCIP49819.2020.9301795

Cite this

Wang, Z. H., Zhao, S., Shen, J., & Lei, Z. (2020). Efficient Light Deep Network for Street Scene Parsing. In 2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020 (pp. 42-45). Article 9301795 (2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/VCIP49819.2020.9301795

Wang, Zhe Hui ; Zhao, Sanyuan ; Shen, Jianbing et al. / Efficient Light Deep Network for Street Scene Parsing. 2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020. Institute of Electrical and Electronics Engineers Inc., 2020. pp. 42-45 (2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020).

@inproceedings{9b5eb53e3f3f401fbcdcf91bf11e7437,

title = "Efficient Light Deep Network for Street Scene Parsing",

abstract = "The semantic segmentation is a dense pixel label pre-diction task, which takes quite a lot of resources and computation cost in most of the time. In our approach, we pay attention to balance the speed and better performance which outperforms the state of the art in speed and accuracy for real-time performance. We come up with the idea of new efficient deep backbone that can extract more semantic details, reduce the computation cost and be easy to deploy at the same time. We call our new backbone as Cascaded Mobile Network, which is proved to be very useful. Our proposed model achieves 72.1 mIOU on the CityScapes val, and 69.5 on CamVid. We achieve good balance between speed and accuracy.",

keywords = "CNN, Deep Learning, Efficient B ackbone, Semantic Segmentation, real-time",

author = "Wang, {Zhe Hui} and Sanyuan Zhao and Jianbing Shen and Zhengchao Lei",

note = "Publisher Copyright: {\textcopyright} 2020 IEEE.; 2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020 ; Conference date: 01-12-2020 Through 04-12-2020",

year = "2020",

month = dec,

day = "1",

doi = "10.1109/VCIP49819.2020.9301795",

language = "English",

series = "2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020",

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

pages = "42--45",

booktitle = "2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020",

address = "United States",

}

Wang, ZH, Zhao, S, Shen, J & Lei, Z 2020, Efficient Light Deep Network for Street Scene Parsing. in 2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020., 9301795, 2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020, Institute of Electrical and Electronics Engineers Inc., pp. 42-45, 2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020, Virtual, Macau, China, 1/12/20. https://doi.org/10.1109/VCIP49819.2020.9301795

Efficient Light Deep Network for Street Scene Parsing. / Wang, Zhe Hui; Zhao, Sanyuan; Shen, Jianbing et al.
2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020. Institute of Electrical and Electronics Engineers Inc., 2020. p. 42-45 9301795 (2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020).

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

TY - GEN

T1 - Efficient Light Deep Network for Street Scene Parsing

AU - Wang, Zhe Hui

AU - Zhao, Sanyuan

AU - Shen, Jianbing

AU - Lei, Zhengchao

PY - 2020/12/1

Y1 - 2020/12/1

N2 - The semantic segmentation is a dense pixel label pre-diction task, which takes quite a lot of resources and computation cost in most of the time. In our approach, we pay attention to balance the speed and better performance which outperforms the state of the art in speed and accuracy for real-time performance. We come up with the idea of new efficient deep backbone that can extract more semantic details, reduce the computation cost and be easy to deploy at the same time. We call our new backbone as Cascaded Mobile Network, which is proved to be very useful. Our proposed model achieves 72.1 mIOU on the CityScapes val, and 69.5 on CamVid. We achieve good balance between speed and accuracy.

AB - The semantic segmentation is a dense pixel label pre-diction task, which takes quite a lot of resources and computation cost in most of the time. In our approach, we pay attention to balance the speed and better performance which outperforms the state of the art in speed and accuracy for real-time performance. We come up with the idea of new efficient deep backbone that can extract more semantic details, reduce the computation cost and be easy to deploy at the same time. We call our new backbone as Cascaded Mobile Network, which is proved to be very useful. Our proposed model achieves 72.1 mIOU on the CityScapes val, and 69.5 on CamVid. We achieve good balance between speed and accuracy.

KW - CNN

KW - Deep Learning

KW - Efficient B ackbone

KW - Semantic Segmentation

KW - real-time

UR - http://www.scopus.com/inward/record.url?scp=85099449332&partnerID=8YFLogxK

U2 - 10.1109/VCIP49819.2020.9301795

DO - 10.1109/VCIP49819.2020.9301795

M3 - Conference contribution

AN - SCOPUS:85099449332

T3 - 2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020

SP - 42

EP - 45

BT - 2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020

Y2 - 1 December 2020 through 4 December 2020

ER -

Wang ZH, Zhao S, Shen J, Lei Z. Efficient Light Deep Network for Street Scene Parsing. In 2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020. Institute of Electrical and Electronics Engineers Inc. 2020. p. 42-45. 9301795. (2020 IEEE International Conference on Visual Communications and Image Processing, VCIP 2020). doi: 10.1109/VCIP49819.2020.9301795

Efficient Light Deep Network for Street Scene Parsing

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this