Generative Neural Networks for Anomaly Detection in Crowded Scenes

Tian Wang; Meina Qiao; Zhiwei Lin; Ce Li; Hichem Snoussi; Zhe Liu; Chang Choi

doi:10.1109/TIFS.2018.2878538

Generative Neural Networks for Anomaly Detection in Crowded Scenes

Tian Wang, Meina Qiao, Zhiwei Lin, Ce Li, Hichem Snoussi, Zhe Liu, Chang Choi^*

^*Corresponding author for this work

School of Mathematics and Physics

Research output: Contribution to journal › Article › peer-review

139 Citations (Scopus)

Abstract

Security surveillance is critical to social harmony and people's peaceful life. It has a great impact on strengthening social stability and life safeguarding. Detecting anomaly timely, effectively and efficiently in video surveillance remains challenging. This paper proposes a new approach, called S²-VAE, for anomaly detection from video data. The S²-VAE consists of two proposed neural networks: a Stacked Fully Connected Variational AutoEncoder (S_F-VAE) and a Skip Convolutional VAE (S_C-VAE). The S_F-VAE is a shallow generative network to obtain a model like Gaussian mixture to fit the distribution of the actual data. The S_C-VAE, as a key component of S²-VAE, is a deep generative network to take advantages of CNN, VAE and skip connections. Both S_F-VAE and S_C-VAE are efficient and effective generative networks and they can achieve better performance for detecting both local abnormal events and global abnormal events. The proposed S²-VAE is evaluated using four public datasets. The experimental results show that the S²-VAE outperforms the state-of-the-art algorithms. The code is available publicly at https://github.com/tianwangbuaa/.

Original language	English
Article number	8513816
Pages (from-to)	1390-1399
Number of pages	10
Journal	IEEE Transactions on Information Forensics and Security
Volume	14
Issue number	5
DOIs	https://doi.org/10.1109/TIFS.2018.2878538
Publication status	Published - 29 Oct 2018

Bibliographical note

Funding Information:
Manuscript received April 16, 2018; revised August 26, 2018; accepted October 13, 2018. Date of publication October 29, 2018; date of current version January 30, 2019. This work was supported in part by the National Natural Science Foundation of China under Grant 61503017, Grant U1435220, Grant 61866022, and Grant 61802180, in part by the Aeronautical Science Foundation of China under Grant 2016ZC51022, in part by the SURECAP CPER Project, in part by the EU Horizon 2020 Research and Innovation Programme under Grant 690238 for DESIREE Project, in part by the UK EPSRC under Grant EP/P031668/1, in part by the BT Ireland Innovation Centre (BTIIC), in part by the Platform CAPSEC funded by Région Champagne-Ardenne and FEDER, and in part by the National Research Foundation of Korea (NRF) Grant funded by the Korea Government (Ministry of Science and ICT) under Grant 2017R1E1A1A01077913. The associate editor coordinating the review of this manuscript and approving it for publication was Dr. Siwei Lyu. (Corresponding author: Chang Choi.) T. Wang and M. Qiao are with the School of Automation Science and Electrical Engineering, Beihang University, Beijing 100083, China (e-mail: [email protected]; [email protected]).

Publisher Copyright:
© 2005-2012 IEEE.

Copyright:
Copyright 2019 Elsevier B.V., All rights reserved.

Keywords

anomaly detection
loss function
Spatio-temporal
variational autoencoder

ASJC Scopus subject areas

Safety, Risk, Reliability and Quality
Computer Networks and Communications

Access to Document

10.1109/TIFS.2018.2878538Licence: Unspecified

Cite this

@article{a477afeb5c9848868484d195a4e1e190,

title = "Generative Neural Networks for Anomaly Detection in Crowded Scenes",

abstract = "Security surveillance is critical to social harmony and people's peaceful life. It has a great impact on strengthening social stability and life safeguarding. Detecting anomaly timely, effectively and efficiently in video surveillance remains challenging. This paper proposes a new approach, called S2-VAE, for anomaly detection from video data. The S2-VAE consists of two proposed neural networks: a Stacked Fully Connected Variational AutoEncoder (SF-VAE) and a Skip Convolutional VAE (SC-VAE). The SF-VAE is a shallow generative network to obtain a model like Gaussian mixture to fit the distribution of the actual data. The SC-VAE, as a key component of S2-VAE, is a deep generative network to take advantages of CNN, VAE and skip connections. Both SF-VAE and SC-VAE are efficient and effective generative networks and they can achieve better performance for detecting both local abnormal events and global abnormal events. The proposed S2-VAE is evaluated using four public datasets. The experimental results show that the S2-VAE outperforms the state-of-the-art algorithms. The code is available publicly at https://github.com/tianwangbuaa/.",

keywords = "anomaly detection, loss function, Spatio-temporal, variational autoencoder",

author = "Tian Wang and Meina Qiao and Zhiwei Lin and Ce Li and Hichem Snoussi and Zhe Liu and Chang Choi",

note = "Funding Information: Manuscript received April 16, 2018; revised August 26, 2018; accepted October 13, 2018. Date of publication October 29, 2018; date of current version January 30, 2019. This work was supported in part by the National Natural Science Foundation of China under Grant 61503017, Grant U1435220, Grant 61866022, and Grant 61802180, in part by the Aeronautical Science Foundation of China under Grant 2016ZC51022, in part by the SURECAP CPER Project, in part by the EU Horizon 2020 Research and Innovation Programme under Grant 690238 for DESIREE Project, in part by the UK EPSRC under Grant EP/P031668/1, in part by the BT Ireland Innovation Centre (BTIIC), in part by the Platform CAPSEC funded by R{\'e}gion Champagne-Ardenne and FEDER, and in part by the National Research Foundation of Korea (NRF) Grant funded by the Korea Government (Ministry of Science and ICT) under Grant 2017R1E1A1A01077913. The associate editor coordinating the review of this manuscript and approving it for publication was Dr. Siwei Lyu. (Corresponding author: Chang Choi.) T. Wang and M. Qiao are with the School of Automation Science and Electrical Engineering, Beihang University, Beijing 100083, China (e-mail: [email protected]; [email protected]). Publisher Copyright: {\textcopyright} 2005-2012 IEEE. Copyright: Copyright 2019 Elsevier B.V., All rights reserved.",

year = "2018",

month = oct,

day = "29",

doi = "10.1109/TIFS.2018.2878538",

language = "English",

volume = "14",

pages = "1390--1399",

journal = "IEEE Transactions on Information Forensics and Security",

issn = "1556-6013",

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

number = "5",

}

TY - JOUR

T1 - Generative Neural Networks for Anomaly Detection in Crowded Scenes

AU - Wang, Tian

AU - Qiao, Meina

AU - Lin, Zhiwei

AU - Li, Ce

AU - Snoussi, Hichem

AU - Liu, Zhe

AU - Choi, Chang

N1 - Funding Information: Manuscript received April 16, 2018; revised August 26, 2018; accepted October 13, 2018. Date of publication October 29, 2018; date of current version January 30, 2019. This work was supported in part by the National Natural Science Foundation of China under Grant 61503017, Grant U1435220, Grant 61866022, and Grant 61802180, in part by the Aeronautical Science Foundation of China under Grant 2016ZC51022, in part by the SURECAP CPER Project, in part by the EU Horizon 2020 Research and Innovation Programme under Grant 690238 for DESIREE Project, in part by the UK EPSRC under Grant EP/P031668/1, in part by the BT Ireland Innovation Centre (BTIIC), in part by the Platform CAPSEC funded by Région Champagne-Ardenne and FEDER, and in part by the National Research Foundation of Korea (NRF) Grant funded by the Korea Government (Ministry of Science and ICT) under Grant 2017R1E1A1A01077913. The associate editor coordinating the review of this manuscript and approving it for publication was Dr. Siwei Lyu. (Corresponding author: Chang Choi.) T. Wang and M. Qiao are with the School of Automation Science and Electrical Engineering, Beihang University, Beijing 100083, China (e-mail: [email protected]; [email protected]). Publisher Copyright: © 2005-2012 IEEE. Copyright: Copyright 2019 Elsevier B.V., All rights reserved.

PY - 2018/10/29

Y1 - 2018/10/29

N2 - Security surveillance is critical to social harmony and people's peaceful life. It has a great impact on strengthening social stability and life safeguarding. Detecting anomaly timely, effectively and efficiently in video surveillance remains challenging. This paper proposes a new approach, called S2-VAE, for anomaly detection from video data. The S2-VAE consists of two proposed neural networks: a Stacked Fully Connected Variational AutoEncoder (SF-VAE) and a Skip Convolutional VAE (SC-VAE). The SF-VAE is a shallow generative network to obtain a model like Gaussian mixture to fit the distribution of the actual data. The SC-VAE, as a key component of S2-VAE, is a deep generative network to take advantages of CNN, VAE and skip connections. Both SF-VAE and SC-VAE are efficient and effective generative networks and they can achieve better performance for detecting both local abnormal events and global abnormal events. The proposed S2-VAE is evaluated using four public datasets. The experimental results show that the S2-VAE outperforms the state-of-the-art algorithms. The code is available publicly at https://github.com/tianwangbuaa/.

AB - Security surveillance is critical to social harmony and people's peaceful life. It has a great impact on strengthening social stability and life safeguarding. Detecting anomaly timely, effectively and efficiently in video surveillance remains challenging. This paper proposes a new approach, called S2-VAE, for anomaly detection from video data. The S2-VAE consists of two proposed neural networks: a Stacked Fully Connected Variational AutoEncoder (SF-VAE) and a Skip Convolutional VAE (SC-VAE). The SF-VAE is a shallow generative network to obtain a model like Gaussian mixture to fit the distribution of the actual data. The SC-VAE, as a key component of S2-VAE, is a deep generative network to take advantages of CNN, VAE and skip connections. Both SF-VAE and SC-VAE are efficient and effective generative networks and they can achieve better performance for detecting both local abnormal events and global abnormal events. The proposed S2-VAE is evaluated using four public datasets. The experimental results show that the S2-VAE outperforms the state-of-the-art algorithms. The code is available publicly at https://github.com/tianwangbuaa/.

KW - anomaly detection

KW - loss function

KW - Spatio-temporal

KW - variational autoencoder

U2 - 10.1109/TIFS.2018.2878538

DO - 10.1109/TIFS.2018.2878538

M3 - Article

AN - SCOPUS:85055697143

SN - 1556-6013

VL - 14

SP - 1390

EP - 1399

JO - IEEE Transactions on Information Forensics and Security

JF - IEEE Transactions on Information Forensics and Security

IS - 5

M1 - 8513816

ER -

Generative Neural Networks for Anomaly Detection in Crowded Scenes

Abstract

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this