Privacy-aware Anomaly Detection using Semantic Segmentation

Michael  Bidstrup; Jacob Velling Dueholm; Kamal Nasrollahi; Thomas B. Moeslund

doi:10.1007/978-3-030-90436-4_9

Privacy-aware Anomaly Detection using Semantic Segmentation

Michael Bidstrup^*, Jacob Velling Dueholm, Kamal Nasrollahi, Thomas B. Moeslund

^*Corresponding author for this work

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

1 Citation (Scopus)

Abstract

This paper sets out to investigate if semantic segmentation
can be used to achieve anonymity in video surveillance while maintaining
the ability to perform anomaly detection. The paper is centered around
finding the best model for segmenting the anomaly detection dataset
UCHK Avenue without semantic ground truth available. To do this, a
series of segmentation models pre-trained on ADE20K and Cityscapes
are evaluated against a custom semantic annotation of selected frames
from Avenue and the segmentation results are compared both quantitatively and qualitatively. The segmented dataset is then tested on a series
of different anomaly detection baselines and the results are compared
both in terms of global and anomaly specific accuracy. When comparing the anomaly detection accuracy for RGB and segmented data it
was found that anonymity in anomaly detection can be achieved at a
small cost in global accuracy but with better accuracy for some specific
anomalies.

Original language	English
Title of host publication	Advances in Visual Computing : 16th International Symposium, ISVC 2021, Virtual Event, October 4-6, 2021, Proceedings, Part II
Editors	George Bebis, Vassilis Athitsos, Tong Yan, Manfred Lau, Frederick Li
Number of pages	123
Volume	13018
Publisher	Springer
Publication date	2021
Pages	110
ISBN (Print)	978-3-030-90435-7
ISBN (Electronic)	978-3-030-90436-4
DOIs	https://doi.org/10.1007/978-3-030-90436-4_9
Publication status	Published - 2021
Event	16th International Symposium on Visual Computing, ISVC 2021 - Virtual Online Duration: 4 Oct 2021 → 6 Oct 2021

Conference

Conference	16th International Symposium on Visual Computing, ISVC 2021
City	Virtual Online
Period	04/10/2021 → 06/10/2021

Series	Lecture Notes in Computer Science
ISSN	0302-9743

Access to Document

10.1007/978-3-030-90436-4_9

AUB Link

Search for the material in Aalborg University Library's search engine

Video Anomaly Detection and Behavior Analysis
Dueholm, J. V.
01/05/2020 → 01/05/2023
Project: PhD Project

Cite this

Bidstrup, M., Dueholm, J. V., Nasrollahi, K., & Moeslund, T. B. (2021). Privacy-aware Anomaly Detection using Semantic Segmentation. In G. Bebis, V. Athitsos, T. Yan, M. Lau, & F. Li (Eds.), Advances in Visual Computing: 16th International Symposium, ISVC 2021, Virtual Event, October 4-6, 2021, Proceedings, Part II (Vol. 13018, pp. 110). Springer. https://doi.org/10.1007/978-3-030-90436-4_9

Bidstrup, Michael ; Dueholm, Jacob Velling ; Nasrollahi, Kamal et al. / Privacy-aware Anomaly Detection using Semantic Segmentation. Advances in Visual Computing: 16th International Symposium, ISVC 2021, Virtual Event, October 4-6, 2021, Proceedings, Part II. editor / George Bebis ; Vassilis Athitsos ; Tong Yan ; Manfred Lau ; Frederick Li. Vol. 13018 Springer, 2021. pp. 110 (Lecture Notes in Computer Science).

@inproceedings{4c1de560e1b642deb79d17185f813191,

title = "Privacy-aware Anomaly Detection using Semantic Segmentation",

abstract = "This paper sets out to investigate if semantic segmentationcan be used to achieve anonymity in video surveillance while maintainingthe ability to perform anomaly detection. The paper is centered aroundfinding the best model for segmenting the anomaly detection datasetUCHK Avenue without semantic ground truth available. To do this, aseries of segmentation models pre-trained on ADE20K and Cityscapesare evaluated against a custom semantic annotation of selected framesfrom Avenue and the segmentation results are compared both quantitatively and qualitatively. The segmented dataset is then tested on a seriesof different anomaly detection baselines and the results are comparedboth in terms of global and anomaly specific accuracy. When comparing the anomaly detection accuracy for RGB and segmented data itwas found that anonymity in anomaly detection can be achieved at asmall cost in global accuracy but with better accuracy for some specificanomalies.",

author = "Michael Bidstrup and Dueholm, {Jacob Velling} and Kamal Nasrollahi and Moeslund, {Thomas B.}",

year = "2021",

doi = "10.1007/978-3-030-90436-4_9",

language = "English",

isbn = "978-3-030-90435-7",

volume = "13018",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "110",

editor = "George Bebis and Vassilis Athitsos and Tong Yan and Manfred Lau and Frederick Li",

booktitle = "Advances in Visual Computing",

address = "Germany",

note = "16th International Symposium on Visual Computing, ISVC 2021 ; Conference date: 04-10-2021 Through 06-10-2021",

}

Bidstrup, M, Dueholm, JV , Nasrollahi, K & Moeslund, TB 2021, Privacy-aware Anomaly Detection using Semantic Segmentation. in G Bebis, V Athitsos, T Yan, M Lau & F Li (eds), Advances in Visual Computing: 16th International Symposium, ISVC 2021, Virtual Event, October 4-6, 2021, Proceedings, Part II. vol. 13018, Springer, Lecture Notes in Computer Science, pp. 110, 16th International Symposium on Visual Computing, ISVC 2021, Virtual Online, 04/10/2021. https://doi.org/10.1007/978-3-030-90436-4_9

Privacy-aware Anomaly Detection using Semantic Segmentation. / Bidstrup, Michael ; Dueholm, Jacob Velling ; Nasrollahi, Kamal et al.
Advances in Visual Computing: 16th International Symposium, ISVC 2021, Virtual Event, October 4-6, 2021, Proceedings, Part II. ed. / George Bebis; Vassilis Athitsos; Tong Yan; Manfred Lau; Frederick Li. Vol. 13018 Springer, 2021. p. 110 (Lecture Notes in Computer Science).

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

TY - GEN

T1 - Privacy-aware Anomaly Detection using Semantic Segmentation

AU - Bidstrup, Michael

AU - Dueholm, Jacob Velling

AU - Nasrollahi, Kamal

AU - Moeslund, Thomas B.

PY - 2021

Y1 - 2021

N2 - This paper sets out to investigate if semantic segmentationcan be used to achieve anonymity in video surveillance while maintainingthe ability to perform anomaly detection. The paper is centered aroundfinding the best model for segmenting the anomaly detection datasetUCHK Avenue without semantic ground truth available. To do this, aseries of segmentation models pre-trained on ADE20K and Cityscapesare evaluated against a custom semantic annotation of selected framesfrom Avenue and the segmentation results are compared both quantitatively and qualitatively. The segmented dataset is then tested on a seriesof different anomaly detection baselines and the results are comparedboth in terms of global and anomaly specific accuracy. When comparing the anomaly detection accuracy for RGB and segmented data itwas found that anonymity in anomaly detection can be achieved at asmall cost in global accuracy but with better accuracy for some specificanomalies.

AB - This paper sets out to investigate if semantic segmentationcan be used to achieve anonymity in video surveillance while maintainingthe ability to perform anomaly detection. The paper is centered aroundfinding the best model for segmenting the anomaly detection datasetUCHK Avenue without semantic ground truth available. To do this, aseries of segmentation models pre-trained on ADE20K and Cityscapesare evaluated against a custom semantic annotation of selected framesfrom Avenue and the segmentation results are compared both quantitatively and qualitatively. The segmented dataset is then tested on a seriesof different anomaly detection baselines and the results are comparedboth in terms of global and anomaly specific accuracy. When comparing the anomaly detection accuracy for RGB and segmented data itwas found that anonymity in anomaly detection can be achieved at asmall cost in global accuracy but with better accuracy for some specificanomalies.

U2 - 10.1007/978-3-030-90436-4_9

DO - 10.1007/978-3-030-90436-4_9

M3 - Article in proceeding

SN - 978-3-030-90435-7

VL - 13018

T3 - Lecture Notes in Computer Science

SP - 110

BT - Advances in Visual Computing

A2 - Bebis, George

A2 - Athitsos, Vassilis

A2 - Yan, Tong

A2 - Lau, Manfred

A2 - Li, Frederick

PB - Springer

T2 - 16th International Symposium on Visual Computing, ISVC 2021

Y2 - 4 October 2021 through 6 October 2021

ER -

Bidstrup M, Dueholm JV , Nasrollahi K , Moeslund TB. Privacy-aware Anomaly Detection using Semantic Segmentation. In Bebis G, Athitsos V, Yan T, Lau M, Li F, editors, Advances in Visual Computing: 16th International Symposium, ISVC 2021, Virtual Event, October 4-6, 2021, Proceedings, Part II. Vol. 13018. Springer. 2021. p. 110. (Lecture Notes in Computer Science). doi: 10.1007/978-3-030-90436-4_9

Privacy-aware Anomaly Detection using Semantic Segmentation

Abstract

Conference

Access to Document

AUB Link

Fingerprint

Projects

Video Anomaly Detection and Behavior Analysis

Cite this