@inproceedings{wang-etal-2025-cmhkf,
title = "{CMHKF}: Cross-Modality Heterogeneous Knowledge Fusion for Weakly Supervised Video Anomaly Detection",
author = "Wang, Guohua and
Song, Shengping and
He, Wuchun and
Zheng, Yongsen",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.acl-long.1524/",
doi = "10.18653/v1/2025.acl-long.1524",
pages = "31594--31607",
ISBN = "979-8-89176-251-0",
abstract = "Weakly supervised video anomaly detection (WSVAD) presents a challenging task focused on detecting frame-level anomalies using only video-level labels. However, existing methods focus mainly on visual modalities, neglecting rich multi-modality information. This paper proposes a novel framework, Cross-Modality Heterogeneous Knowledge Fusion (CMHKF), that integrates cross-modality knowledge from video, audio, and text to improve anomaly detection and localization. To achieve adaptive cross-modality heterogeneous knowledge learning, we designed two components: Cross-Modality Video-Text Knowledge Alignment (CVKA) and Audio Modality Feature Adaptive Extraction (AFAE). They extract and aggregate features by exploring inter-modality correlations. By leveraging abundant cross-modality knowledge, our approach improves the discrimination between normal and anomalous segments. Extensive experiments on XD-Violence show our method significantly enhances accuracy and robustness in both coarse-grained and fine-grained anomaly detection."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2025-cmhkf">
<titleInfo>
<title>CMHKF: Cross-Modality Heterogeneous Knowledge Fusion for Weakly Supervised Video Anomaly Detection</title>
</titleInfo>
<name type="personal">
<namePart type="given">Guohua</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shengping</namePart>
<namePart type="family">Song</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wuchun</namePart>
<namePart type="family">He</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yongsen</namePart>
<namePart type="family">Zheng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wanxiang</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joyce</namePart>
<namePart type="family">Nabende</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Shutova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-251-0</identifier>
</relatedItem>
<abstract>Weakly supervised video anomaly detection (WSVAD) presents a challenging task focused on detecting frame-level anomalies using only video-level labels. However, existing methods focus mainly on visual modalities, neglecting rich multi-modality information. This paper proposes a novel framework, Cross-Modality Heterogeneous Knowledge Fusion (CMHKF), that integrates cross-modality knowledge from video, audio, and text to improve anomaly detection and localization. To achieve adaptive cross-modality heterogeneous knowledge learning, we designed two components: Cross-Modality Video-Text Knowledge Alignment (CVKA) and Audio Modality Feature Adaptive Extraction (AFAE). They extract and aggregate features by exploring inter-modality correlations. By leveraging abundant cross-modality knowledge, our approach improves the discrimination between normal and anomalous segments. Extensive experiments on XD-Violence show our method significantly enhances accuracy and robustness in both coarse-grained and fine-grained anomaly detection.</abstract>
<identifier type="citekey">wang-etal-2025-cmhkf</identifier>
<identifier type="doi">10.18653/v1/2025.acl-long.1524</identifier>
<location>
<url>https://aclanthology.org/2025.acl-long.1524/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>31594</start>
<end>31607</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T CMHKF: Cross-Modality Heterogeneous Knowledge Fusion for Weakly Supervised Video Anomaly Detection
%A Wang, Guohua
%A Song, Shengping
%A He, Wuchun
%A Zheng, Yongsen
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-251-0
%F wang-etal-2025-cmhkf
%X Weakly supervised video anomaly detection (WSVAD) presents a challenging task focused on detecting frame-level anomalies using only video-level labels. However, existing methods focus mainly on visual modalities, neglecting rich multi-modality information. This paper proposes a novel framework, Cross-Modality Heterogeneous Knowledge Fusion (CMHKF), that integrates cross-modality knowledge from video, audio, and text to improve anomaly detection and localization. To achieve adaptive cross-modality heterogeneous knowledge learning, we designed two components: Cross-Modality Video-Text Knowledge Alignment (CVKA) and Audio Modality Feature Adaptive Extraction (AFAE). They extract and aggregate features by exploring inter-modality correlations. By leveraging abundant cross-modality knowledge, our approach improves the discrimination between normal and anomalous segments. Extensive experiments on XD-Violence show our method significantly enhances accuracy and robustness in both coarse-grained and fine-grained anomaly detection.
%R 10.18653/v1/2025.acl-long.1524
%U https://aclanthology.org/2025.acl-long.1524/
%U https://doi.org/10.18653/v1/2025.acl-long.1524
%P 31594-31607
Markdown (Informal)
[CMHKF: Cross-Modality Heterogeneous Knowledge Fusion for Weakly Supervised Video Anomaly Detection](https://aclanthology.org/2025.acl-long.1524/) (Wang et al., ACL 2025)
ACL