TeG: Temporal-Granularity Method for Anomaly Detection with Attention in Smart City Surveillance

Erkut Akdag; Egor Bondarev; Peter H.N. de With

doi:10.1109/VCIP63160.2024.10849819

TeG: Temporal-Granularity Method for Anomaly Detection with Attention in Smart City Surveillance

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

2 Citations (Scopus)

14 Downloads (Pure)

Abstract

Anomaly detection in video surveillance has recently gained interest from the research community. Temporal duration of anomalies vary within video streams, leading to complications in learning the temporal dynamics of specific events. This paper presents a temporal-granularity method for an anomaly detection model (TeG) in real-world surveillance, combining spatio-temporal features at different time-scales. The TeG model employs multi-head cross-attention (MCA) blocks and multi-head self-attention (MSA) blocks for this purpose. Additionally, we extend the UCF-Crime dataset with new anomaly types relevant to Smart City research project. The TeG model is deployed and validated in a city surveillance system, achieving successful real-time results in industrial settings.

Original language	English
Title of host publication	2024 IEEE International Conference on Visual Communications and Image Processing, VCIP 2024
Publisher	Institute of Electrical and Electronics Engineers
Number of pages	6
ISBN (Electronic)	979-8-3315-2954-3
DOIs	https://doi.org/10.1109/VCIP63160.2024.10849819
Publication status	Published - 27 Jan 2025
Event	2024 IEEE International Conference on Visual Communications and Image Processing, IEEE VCIP 2024 - Tokyo, Japan Duration: 8 Dec 2024 → 11 Dec 2024

Conference

Conference	2024 IEEE International Conference on Visual Communications and Image Processing, IEEE VCIP 2024
Abbreviated title	IEEE VCIP 2024
Country/Territory	Japan
City	Tokyo
Period	8/12/24 → 11/12/24

Funding

This work was supported by the European ITEA SMART Mobility project on intelligent traffic flow systems.

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 11 Sustainable Cities and Communities
SDG 16 Peace, Justice and Strong Institutions

Keywords

abnormal behaviour
attention
computer vision
surveillance
temporal granularity

Access to Document

10.1109/VCIP63160.2024.10849819

pdfFinal published version, 2.48 MBLicence: TAVERNE

Cite this

@inproceedings{9fd50417528d4e0f8b098e924f1023af,

title = "TeG: Temporal-Granularity Method for Anomaly Detection with Attention in Smart City Surveillance",

abstract = "Anomaly detection in video surveillance has recently gained interest from the research community. Temporal duration of anomalies vary within video streams, leading to complications in learning the temporal dynamics of specific events. This paper presents a temporal-granularity method for an anomaly detection model (TeG) in real-world surveillance, combining spatio-temporal features at different time-scales. The TeG model employs multi-head cross-attention (MCA) blocks and multi-head self-attention (MSA) blocks for this purpose. Additionally, we extend the UCF-Crime dataset with new anomaly types relevant to Smart City research project. The TeG model is deployed and validated in a city surveillance system, achieving successful real-time results in industrial settings.",

keywords = "abnormal behaviour, attention, computer vision, surveillance, temporal granularity",

author = "Erkut Akdag and Egor Bondarev and \{de With\}, \{Peter H.N.\}",

year = "2025",

month = jan,

day = "27",

doi = "10.1109/VCIP63160.2024.10849819",

language = "English",

booktitle = "2024 IEEE International Conference on Visual Communications and Image Processing, VCIP 2024",

publisher = "Institute of Electrical and Electronics Engineers",

address = "United States",

note = " 2024 IEEE International Conference on Visual Communications and Image Processing, IEEE VCIP 2024, IEEE VCIP 2024 ; Conference date: 08-12-2024 Through 11-12-2024",

}

Akdag, E , Bondarev, E & de With, PHN 2025, TeG: Temporal-Granularity Method for Anomaly Detection with Attention in Smart City Surveillance. in 2024 IEEE International Conference on Visual Communications and Image Processing, VCIP 2024., 10849819, Institute of Electrical and Electronics Engineers, 2024 IEEE International Conference on Visual Communications and Image Processing, IEEE VCIP 2024, Tokyo, Japan, 8/12/24. https://doi.org/10.1109/VCIP63160.2024.10849819

TeG: Temporal-Granularity Method for Anomaly Detection with Attention in Smart City Surveillance. / Akdag, Erkut ; Bondarev, Egor ; de With, Peter H.N.
2024 IEEE International Conference on Visual Communications and Image Processing, VCIP 2024. Institute of Electrical and Electronics Engineers, 2025. 10849819.

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

TY - GEN

T1 - TeG

T2 - 2024 IEEE International Conference on Visual Communications and Image Processing, IEEE VCIP 2024

AU - Akdag, Erkut

AU - Bondarev, Egor

AU - de With, Peter H.N.

PY - 2025/1/27

Y1 - 2025/1/27

N2 - Anomaly detection in video surveillance has recently gained interest from the research community. Temporal duration of anomalies vary within video streams, leading to complications in learning the temporal dynamics of specific events. This paper presents a temporal-granularity method for an anomaly detection model (TeG) in real-world surveillance, combining spatio-temporal features at different time-scales. The TeG model employs multi-head cross-attention (MCA) blocks and multi-head self-attention (MSA) blocks for this purpose. Additionally, we extend the UCF-Crime dataset with new anomaly types relevant to Smart City research project. The TeG model is deployed and validated in a city surveillance system, achieving successful real-time results in industrial settings.

AB - Anomaly detection in video surveillance has recently gained interest from the research community. Temporal duration of anomalies vary within video streams, leading to complications in learning the temporal dynamics of specific events. This paper presents a temporal-granularity method for an anomaly detection model (TeG) in real-world surveillance, combining spatio-temporal features at different time-scales. The TeG model employs multi-head cross-attention (MCA) blocks and multi-head self-attention (MSA) blocks for this purpose. Additionally, we extend the UCF-Crime dataset with new anomaly types relevant to Smart City research project. The TeG model is deployed and validated in a city surveillance system, achieving successful real-time results in industrial settings.

KW - abnormal behaviour

KW - attention

KW - computer vision

KW - surveillance

KW - temporal granularity

UR - https://www.scopus.com/pages/publications/85218197678

U2 - 10.1109/VCIP63160.2024.10849819

DO - 10.1109/VCIP63160.2024.10849819

M3 - Conference contribution

AN - SCOPUS:85218197678

BT - 2024 IEEE International Conference on Visual Communications and Image Processing, VCIP 2024

PB - Institute of Electrical and Electronics Engineers

Y2 - 8 December 2024 through 11 December 2024

ER -

TeG: Temporal-Granularity Method for Anomaly Detection with Attention in Smart City Surveillance

Abstract

Conference

Funding

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Cite this