Enhanced Video Anomaly Detection Through Dual Triplet Contrastive Loss for Hard Sample Discrimination

Learning discriminative features between abnormal and normal instances is crucial for video anomaly detection within the multiple instance learning framework. Existing methods primarily focus on instances with the highest anomaly scores, neglecting the identification and differentiation of hard samp...

Full description

Saved in:
Bibliographic Details
Main Authors: Chunxiang Niu, Siyu Meng, Rong Wang
Format: Article
Language:English
Published: MDPI AG 2025-06-01
Series:Entropy
Subjects:
Online Access:https://www.mdpi.com/1099-4300/27/7/655
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849246551086465024
author Chunxiang Niu
Siyu Meng
Rong Wang
author_facet Chunxiang Niu
Siyu Meng
Rong Wang
author_sort Chunxiang Niu
collection DOAJ
description Learning discriminative features between abnormal and normal instances is crucial for video anomaly detection within the multiple instance learning framework. Existing methods primarily focus on instances with the highest anomaly scores, neglecting the identification and differentiation of hard samples, leading to misjudgments and high false alarm rates. To address these challenges, we propose a dual triplet contrastive loss strategy. This approach employs dual memory units to extract four key feature categories: hard samples, negative samples, positive samples, and anchor samples. Contrastive loss is utilized to constrain the distance between hard samples and other samples, enabling accurate identification of hard samples and enhancing the discriminative ability of hard samples and abnormal features. Additionally, a multi-scale feature perception module is designed to capture feature information at different levels, while an adaptive global–local feature fusion module constructs complementary feature enhancement through feature fusion. Experimental results demonstrate the effectiveness of our method, achieving AUC scores of 87.16% on the UCF-Crime dataset and AP scores of 83.47% on the XD-Violence dataset.
format Article
id doaj-art-3d5fff80721842e8bca9784c576a86e4
institution Kabale University
issn 1099-4300
language English
publishDate 2025-06-01
publisher MDPI AG
record_format Article
series Entropy
spelling doaj-art-3d5fff80721842e8bca9784c576a86e42025-08-20T03:58:27ZengMDPI AGEntropy1099-43002025-06-0127765510.3390/e27070655Enhanced Video Anomaly Detection Through Dual Triplet Contrastive Loss for Hard Sample DiscriminationChunxiang Niu0Siyu Meng1Rong Wang2College of Information and Cyber Security, People’s Public Security University of China, Beijing 100038, ChinaCollege of Information and Cyber Security, People’s Public Security University of China, Beijing 100038, ChinaCollege of Information and Cyber Security, People’s Public Security University of China, Beijing 100038, ChinaLearning discriminative features between abnormal and normal instances is crucial for video anomaly detection within the multiple instance learning framework. Existing methods primarily focus on instances with the highest anomaly scores, neglecting the identification and differentiation of hard samples, leading to misjudgments and high false alarm rates. To address these challenges, we propose a dual triplet contrastive loss strategy. This approach employs dual memory units to extract four key feature categories: hard samples, negative samples, positive samples, and anchor samples. Contrastive loss is utilized to constrain the distance between hard samples and other samples, enabling accurate identification of hard samples and enhancing the discriminative ability of hard samples and abnormal features. Additionally, a multi-scale feature perception module is designed to capture feature information at different levels, while an adaptive global–local feature fusion module constructs complementary feature enhancement through feature fusion. Experimental results demonstrate the effectiveness of our method, achieving AUC scores of 87.16% on the UCF-Crime dataset and AP scores of 83.47% on the XD-Violence dataset.https://www.mdpi.com/1099-4300/27/7/655abnormal behavior detectionmultiple instance learninghard instancecontrastive loss functionmulti-scale feature
spellingShingle Chunxiang Niu
Siyu Meng
Rong Wang
Enhanced Video Anomaly Detection Through Dual Triplet Contrastive Loss for Hard Sample Discrimination
Entropy
abnormal behavior detection
multiple instance learning
hard instance
contrastive loss function
multi-scale feature
title Enhanced Video Anomaly Detection Through Dual Triplet Contrastive Loss for Hard Sample Discrimination
title_full Enhanced Video Anomaly Detection Through Dual Triplet Contrastive Loss for Hard Sample Discrimination
title_fullStr Enhanced Video Anomaly Detection Through Dual Triplet Contrastive Loss for Hard Sample Discrimination
title_full_unstemmed Enhanced Video Anomaly Detection Through Dual Triplet Contrastive Loss for Hard Sample Discrimination
title_short Enhanced Video Anomaly Detection Through Dual Triplet Contrastive Loss for Hard Sample Discrimination
title_sort enhanced video anomaly detection through dual triplet contrastive loss for hard sample discrimination
topic abnormal behavior detection
multiple instance learning
hard instance
contrastive loss function
multi-scale feature
url https://www.mdpi.com/1099-4300/27/7/655
work_keys_str_mv AT chunxiangniu enhancedvideoanomalydetectionthroughdualtripletcontrastivelossforhardsamplediscrimination
AT siyumeng enhancedvideoanomalydetectionthroughdualtripletcontrastivelossforhardsamplediscrimination
AT rongwang enhancedvideoanomalydetectionthroughdualtripletcontrastivelossforhardsamplediscrimination