EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Learning Fine-Grained Video Anomaly Detection from Normal Videos

Ruqin Wang, Yasumasa Tamura iD, Masahito Yamamoto iD

Sensors · 07.07.2026 · 26 (13) · 4314

DOI: 10.3390/s26134314 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 07.07.2026. Verfasst von Ruqin Wang und 2 weiteren Autor:innen. Beteiligte Einrichtung: Graduate School of Information Science and Technology, Hokkaido University, Sapporo 001-0021, Japan. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 38 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Japan Society for the Promotion of Science, JST SPRING, Hokkaido University.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Video anomaly detection (VAD) aims to identify abnormal events in videos. Due to the lack of high-quality training data with detailed annotations, current VAD methods can only produce video-level predictions. To remedy this, several methods attempt to synthesize pseudo video anomalies. However, these methods suffer from low realism and coarse annotations, which limits their performance in real-world scenarios. In this paper, we propose a framework for unsupervised anomaly video generation from solely normal videos, leveraging VLMs to generate structured textual descriptions of anomalies conditioned on the perception of this video. Then, abnormal segments are synthesized using VLMs based on the synthetic textual descriptions. As our framework is highly controllable, video-level and region-level labels can be obtained to provide fine-grained annotations. On top of the synthetic data, we develop a fine-grained VAD network to simultaneously produce video-level, frame-level, and region-level predictions. Experiments show that our method achieves remarkable fine-grained VAD performance.

Schlagwörter

Stichwörter der Autor:innen

video anomaly detectionvision–language modelweak supervision

Autor:innen und Institutionen

  • Graduate School of Information Science and Technology, Hokkaido University, Sapporo 001-0021, Japan.
  • Faculty of Information Science and Technology, Hokkaido University, Sapporo 060-0814, Japan.

ORCID-Profile: Yasumasa Tamura, Masahito Yamamoto

Förderung

  • Japan Society for the Promotion of Science
  • JST SPRING
  • Hokkaido University

Zitieren

APA

Wang, R., Tamura, Y., & Yamamoto, M. (2026). Learning Fine-Grained Video Anomaly Detection from Normal Videos. Sensors, 26(13), 4314. https://doi.org/10.3390/s26134314

Vancouver

Wang R, Tamura Y, Yamamoto M. Learning Fine-Grained Video Anomaly Detection from Normal Videos. Sensors. 2026;26(13):4314. doi:10.3390/s26134314

BibTeX
@article{Wang2026,
  author = {Wang, Ruqin and Tamura, Yasumasa and Yamamoto, Masahito},
  title = {Learning Fine-Grained Video Anomaly Detection from Normal Videos},
  journal = {Sensors},
  year = {2026},
  volume = {26},
  number = {13},
  pages = {4314},
  doi = {10.3390/s26134314}
}
Kostenlos starten →