EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

CNN-ViT Supported Weakly-Supervised Video Segment Level Anomaly Detection

Md Haidar Sharif iD, Lei Jiao iD, Christian W Omlin

Sensors · 07.09.2023 · 23 (18) · 7734

DOI: 10.3390/s23187734 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 07.09.2023. Verfasst von Md Haidar Sharif und 2 weiteren Autor:innen. Beteiligte Einrichtung: Department of ICT, University of Agder, 4630 Kristiansand, Norway. Die Arbeit wurde bisher 22-mal zitiert (Crossref) und nennt 73 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: The Research Council of Norway, Research Council of Norway.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Video anomaly event detection (VAED) is one of the key technologies in computer vision for smart surveillance systems. With the advent of deep learning, contemporary advances in VAED have achieved substantial success. Recently, weakly supervised VAED (WVAED) has become a popular VAED technical route of research. WVAED methods do not depend on a supplementary self-supervised substitute task, yet they can assess anomaly scores straightway. However, the performance of WVAED methods depends on pretrained feature extractors. In this paper, we first address taking advantage of two pretrained feature extractors for CNN (e.g., C3D and I3D) and ViT (e.g., CLIP), for effectively extracting discerning representations. We then consider long-range and short-range temporal dependencies and put forward video snippets of interest by leveraging our proposed temporal self-attention network (TSAN). We design a multiple instance learning (MIL)-based generalized architecture named CNN-ViT-TSAN, by using CNN- and/or ViT-extracted features and TSAN to specify a series of models for the WVAED problem. Experimental results on publicly available popular crowd datasets demonstrated the effectiveness of our CNN-ViT-TSAN.

Schlagwörter

Stichwörter der Autor:innen

Mahalanobis distanceattentionconvolutional neural network (CNN)multiple instance learning (MIL)vision transformer (ViT)weakly supervised video anomaly event detection

Autor:innen und Institutionen

  • Department of ICT, University of Agder, 4630 Kristiansand, Norway.

ORCID-Profile: Md Haidar Sharif, Lei Jiao

Förderung

  • The Research Council of Norway
  • Research Council of Norway

Zitieren

APA

Sharif, M. H., Jiao, L., & Omlin, C. W. (2023). CNN-ViT Supported Weakly-Supervised Video Segment Level Anomaly Detection. Sensors, 23(18), 7734. https://doi.org/10.3390/s23187734

Vancouver

Sharif MH, Jiao L, Omlin CW. CNN-ViT Supported Weakly-Supervised Video Segment Level Anomaly Detection. Sensors. 2023;23(18):7734. doi:10.3390/s23187734

BibTeX
@article{Sharif2023,
  author = {Sharif, Md Haidar and Jiao, Lei and Omlin, Christian W},
  title = {CNN-ViT Supported Weakly-Supervised Video Segment Level Anomaly Detection},
  journal = {Sensors},
  year = {2023},
  volume = {23},
  number = {18},
  pages = {7734},
  doi = {10.3390/s23187734}
}
Kostenlos starten →