A novel reconstruction-based video anomaly detection with idempotent generative network

Video anomaly detection (VAD) is vital in intelligent security for public safety. Reconstruction-based VAD has received increasing research attention, but faces challenges such as missing anomalies for the reconstruction error as a criterion, and information loss when suppressing anomalous data, exi...

Full description

Saved in:
Bibliographic Details
Main Authors: Wenmin Dong, Lifeng Zhang, Wenjuan Shi, Xiangwei Zheng, Yuang Zhang
Format: Article
Language:English
Published: Elsevier 2025-06-01
Series:Alexandria Engineering Journal
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S1110016825004144
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849334919243759616
author Wenmin Dong
Lifeng Zhang
Wenjuan Shi
Xiangwei Zheng
Yuang Zhang
author_facet Wenmin Dong
Lifeng Zhang
Wenjuan Shi
Xiangwei Zheng
Yuang Zhang
author_sort Wenmin Dong
collection DOAJ
description Video anomaly detection (VAD) is vital in intelligent security for public safety. Reconstruction-based VAD has received increasing research attention, but faces challenges such as missing anomalies for the reconstruction error as a criterion, and information loss when suppressing anomalous data, existing methods also struggle to detect unseen anomalies. We propose a novel reconstruction-based video anomaly detection with idempotent generative network (RVADIGN), which is composed of the novel reconstruction module namely PSVAE and an idempotent loss term (IGN). Specifically, video frames are reconstructed within PSVAE. During this process, skip connections are established between the encoder and decoder to enhance contextual understanding. Finite Scalar Quantization (FSQ) layer is designed to discretize the encoder’s output, preserving key discriminative features. Meanwhile, the Pyramid Deformation Module (PDM), as an integral part of PSVAE, computes offset maps of original video frames for anomaly detection supplementation. Alongside PSVAE, idempotence is introduced as a regularity term, which projects the anomaly information back to the estimated manifolds of the target distribution, improves the adaptability and stability of the reconstruction method in different videos. Extensive experimental results demonstrate that our method outperforms other state-of-the-art VAD methods, achieving 99.03%, 92.40%, and 77.20% AUC on UCSD Ped2, CUHK Avenue, and ShanghaiTech, respectively.
format Article
id doaj-art-2e02cc25adfe471fbb245a1ccd4e6d79
institution Kabale University
issn 1110-0168
language English
publishDate 2025-06-01
publisher Elsevier
record_format Article
series Alexandria Engineering Journal
spelling doaj-art-2e02cc25adfe471fbb245a1ccd4e6d792025-08-20T03:45:27ZengElsevierAlexandria Engineering Journal1110-01682025-06-0112451352510.1016/j.aej.2025.03.106A novel reconstruction-based video anomaly detection with idempotent generative networkWenmin Dong0Lifeng Zhang1Wenjuan Shi2Xiangwei Zheng3Yuang Zhang4School of Information Science and Engineering, Shandong Normal University, Jinan, 250358, China; Shandong Provincial Key Laboratory for Distributed Computer Software Novel Technology, Jinan, 250358, ChinaSchool of Information Science and Engineering, Shandong Normal University, Jinan, 250358, China; Shandong Provincial Key Laboratory for Distributed Computer Software Novel Technology, Jinan, 250358, ChinaSchool of Information Science and Engineering, Shandong Normal University, Jinan, 250358, China; Shandong Provincial Key Laboratory for Distributed Computer Software Novel Technology, Jinan, 250358, ChinaCorresponding author at: School of Information Science and Engineering, Shandong Normal University, Jinan, 250358, China.; School of Information Science and Engineering, Shandong Normal University, Jinan, 250358, China; Shandong Provincial Key Laboratory for Distributed Computer Software Novel Technology, Jinan, 250358, ChinaSchool of Information Science and Engineering, Shandong Normal University, Jinan, 250358, China; Shandong Provincial Key Laboratory for Distributed Computer Software Novel Technology, Jinan, 250358, ChinaVideo anomaly detection (VAD) is vital in intelligent security for public safety. Reconstruction-based VAD has received increasing research attention, but faces challenges such as missing anomalies for the reconstruction error as a criterion, and information loss when suppressing anomalous data, existing methods also struggle to detect unseen anomalies. We propose a novel reconstruction-based video anomaly detection with idempotent generative network (RVADIGN), which is composed of the novel reconstruction module namely PSVAE and an idempotent loss term (IGN). Specifically, video frames are reconstructed within PSVAE. During this process, skip connections are established between the encoder and decoder to enhance contextual understanding. Finite Scalar Quantization (FSQ) layer is designed to discretize the encoder’s output, preserving key discriminative features. Meanwhile, the Pyramid Deformation Module (PDM), as an integral part of PSVAE, computes offset maps of original video frames for anomaly detection supplementation. Alongside PSVAE, idempotence is introduced as a regularity term, which projects the anomaly information back to the estimated manifolds of the target distribution, improves the adaptability and stability of the reconstruction method in different videos. Extensive experimental results demonstrate that our method outperforms other state-of-the-art VAD methods, achieving 99.03%, 92.40%, and 77.20% AUC on UCSD Ped2, CUHK Avenue, and ShanghaiTech, respectively.http://www.sciencedirect.com/science/article/pii/S1110016825004144Video anomaly detectionUnsupervised learningIdempotent generative networksQuantitative layerReconstruction supplementation
spellingShingle Wenmin Dong
Lifeng Zhang
Wenjuan Shi
Xiangwei Zheng
Yuang Zhang
A novel reconstruction-based video anomaly detection with idempotent generative network
Alexandria Engineering Journal
Video anomaly detection
Unsupervised learning
Idempotent generative networks
Quantitative layer
Reconstruction supplementation
title A novel reconstruction-based video anomaly detection with idempotent generative network
title_full A novel reconstruction-based video anomaly detection with idempotent generative network
title_fullStr A novel reconstruction-based video anomaly detection with idempotent generative network
title_full_unstemmed A novel reconstruction-based video anomaly detection with idempotent generative network
title_short A novel reconstruction-based video anomaly detection with idempotent generative network
title_sort novel reconstruction based video anomaly detection with idempotent generative network
topic Video anomaly detection
Unsupervised learning
Idempotent generative networks
Quantitative layer
Reconstruction supplementation
url http://www.sciencedirect.com/science/article/pii/S1110016825004144
work_keys_str_mv AT wenmindong anovelreconstructionbasedvideoanomalydetectionwithidempotentgenerativenetwork
AT lifengzhang anovelreconstructionbasedvideoanomalydetectionwithidempotentgenerativenetwork
AT wenjuanshi anovelreconstructionbasedvideoanomalydetectionwithidempotentgenerativenetwork
AT xiangweizheng anovelreconstructionbasedvideoanomalydetectionwithidempotentgenerativenetwork
AT yuangzhang anovelreconstructionbasedvideoanomalydetectionwithidempotentgenerativenetwork
AT wenmindong novelreconstructionbasedvideoanomalydetectionwithidempotentgenerativenetwork
AT lifengzhang novelreconstructionbasedvideoanomalydetectionwithidempotentgenerativenetwork
AT wenjuanshi novelreconstructionbasedvideoanomalydetectionwithidempotentgenerativenetwork
AT xiangweizheng novelreconstructionbasedvideoanomalydetectionwithidempotentgenerativenetwork
AT yuangzhang novelreconstructionbasedvideoanomalydetectionwithidempotentgenerativenetwork