Depth driven people counting using deep region proposal network

D. Song; Y. Qiao; A. Corbetta

doi:10.1109/ICInfA.2017.8078944

Depth driven people counting using deep region proposal network

D. Song, Y. Qiao, A. Corbetta

Fluids and Flows

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

28 Citations (Scopus)

Abstract

People counting is a crucial subject in video surveillance application. Factors such as severe occlusions, scene perspective distortions in real application scenario make this task challenging. In this paper, we carefully designed a deep detection framework based on depth information for people counting in crowded environments. Our system performs head detection on depth images collected by an overhead vertical Kinect sensor. To the best of our knowledge, this is the first attempt to use the deep convolutional neural networks on depth images for people counting. We explored the impact of the number and quality of RPN positive anchors on the performance of Faster R-CNN and proposed a solution. Our method is very simple but effective, not only showing promising results but also efficiency as it runs in real-time at a frame rate of about 110 frames per second on a GPU.

Original language	English
Title of host publication	2017 IEEE International Conference on Information and Automation (ICIA), 18-20 July 2017, Macau, China
Place of Publication	Piscataway
Publisher	Institute of Electrical and Electronics Engineers
Pages	416-421
Number of pages	6
ISBN (Electronic)	978-1-5386-3154-6
ISBN (Print)	978-1-5386-3155-3
DOIs	https://doi.org/10.1109/ICInfA.2017.8078944
Publication status	Published - 20 Oct 2017
Event	2017 IEEE International Conference on Information and Automation, (ICIA 2017) - Macau, China Duration: 18 Jul 2017 → 20 Jul 2017 http://2017.ieee-icia.org/

Conference

Conference	2017 IEEE International Conference on Information and Automation, (ICIA 2017)
Abbreviated title	ICIA 2017
Country/Territory	China
City	Macau
Period	18/07/17 → 20/07/17
Internet address	http://2017.ieee-icia.org/

Funding

* This work is partly supported by External Cooperation Program of BIC Chinese Academy of Sciences Grant #172644KYSB20150019, and Shenzhen Research Program Grant #JSGG20150925164740726, Grant #KQCX2015033117354153. This work is part of the JSTP research programme ``Vision driven visitor behaviour analysis and crowd management'' with project number 341-10-001, which is financed by the Netherlands Organisation for Scientific Research (NWO).

Keywords

Convolutional Neural Network
Depth image
Head detection
People counting

Access to Document

10.1109/ICInfA.2017.8078944

Cite this

@inproceedings{62755f2f4b794735999deea0a5290a44,

title = "Depth driven people counting using deep region proposal network",

abstract = "People counting is a crucial subject in video surveillance application. Factors such as severe occlusions, scene perspective distortions in real application scenario make this task challenging. In this paper, we carefully designed a deep detection framework based on depth information for people counting in crowded environments. Our system performs head detection on depth images collected by an overhead vertical Kinect sensor. To the best of our knowledge, this is the first attempt to use the deep convolutional neural networks on depth images for people counting. We explored the impact of the number and quality of RPN positive anchors on the performance of Faster R-CNN and proposed a solution. Our method is very simple but effective, not only showing promising results but also efficiency as it runs in real-time at a frame rate of about 110 frames per second on a GPU.",

keywords = "Convolutional Neural Network, Depth image, Head detection, People counting",

author = "D. Song and Y. Qiao and A. Corbetta",

year = "2017",

month = oct,

day = "20",

doi = "10.1109/ICInfA.2017.8078944",

language = "English",

isbn = "978-1-5386-3155-3 ",

pages = "416--421",

booktitle = "2017 IEEE International Conference on Information and Automation (ICIA), 18-20 July 2017, Macau, China",

publisher = "Institute of Electrical and Electronics Engineers",

address = "United States",

note = "2017 IEEE International Conference on Information and Automation, (ICIA 2017), ICIA 2017 ; Conference date: 18-07-2017 Through 20-07-2017",

url = "http://2017.ieee-icia.org/",

}

Song, D, Qiao, Y & Corbetta, A 2017, Depth driven people counting using deep region proposal network. in 2017 IEEE International Conference on Information and Automation (ICIA), 18-20 July 2017, Macau, China., 8078944, Institute of Electrical and Electronics Engineers, Piscataway, pp. 416-421, 2017 IEEE International Conference on Information and Automation, (ICIA 2017), Macau, China, 18/07/17. https://doi.org/10.1109/ICInfA.2017.8078944

Depth driven people counting using deep region proposal network. / Song, D.; Qiao, Y.; Corbetta, A.
2017 IEEE International Conference on Information and Automation (ICIA), 18-20 July 2017, Macau, China. Piscataway: Institute of Electrical and Electronics Engineers, 2017. p. 416-421 8078944.

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

TY - GEN

T1 - Depth driven people counting using deep region proposal network

AU - Song, D.

AU - Qiao, Y.

AU - Corbetta, A.

PY - 2017/10/20

Y1 - 2017/10/20

N2 - People counting is a crucial subject in video surveillance application. Factors such as severe occlusions, scene perspective distortions in real application scenario make this task challenging. In this paper, we carefully designed a deep detection framework based on depth information for people counting in crowded environments. Our system performs head detection on depth images collected by an overhead vertical Kinect sensor. To the best of our knowledge, this is the first attempt to use the deep convolutional neural networks on depth images for people counting. We explored the impact of the number and quality of RPN positive anchors on the performance of Faster R-CNN and proposed a solution. Our method is very simple but effective, not only showing promising results but also efficiency as it runs in real-time at a frame rate of about 110 frames per second on a GPU.

AB - People counting is a crucial subject in video surveillance application. Factors such as severe occlusions, scene perspective distortions in real application scenario make this task challenging. In this paper, we carefully designed a deep detection framework based on depth information for people counting in crowded environments. Our system performs head detection on depth images collected by an overhead vertical Kinect sensor. To the best of our knowledge, this is the first attempt to use the deep convolutional neural networks on depth images for people counting. We explored the impact of the number and quality of RPN positive anchors on the performance of Faster R-CNN and proposed a solution. Our method is very simple but effective, not only showing promising results but also efficiency as it runs in real-time at a frame rate of about 110 frames per second on a GPU.

KW - Convolutional Neural Network

KW - Depth image

KW - Head detection

KW - People counting

UR - http://www.scopus.com/inward/record.url?scp=85039963127&partnerID=8YFLogxK

U2 - 10.1109/ICInfA.2017.8078944

DO - 10.1109/ICInfA.2017.8078944

M3 - Conference contribution

AN - SCOPUS:85039963127

SN - 978-1-5386-3155-3

SP - 416

EP - 421

BT - 2017 IEEE International Conference on Information and Automation (ICIA), 18-20 July 2017, Macau, China

PB - Institute of Electrical and Electronics Engineers

CY - Piscataway

T2 - 2017 IEEE International Conference on Information and Automation, (ICIA 2017)

Y2 - 18 July 2017 through 20 July 2017

ER -

Depth driven people counting using deep region proposal network

Abstract

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this