EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

SETJiP: Spatial and Extra Temporal Jigsaw Puzzles for Video Anomaly Detection

Liheng Shen iD, Tetsu Matsukawa iD, Einoshin Suzuki iD

Sensors · 05.05.2026 · 26 (9) · 2889

DOI: 10.3390/s26092889 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 05.05.2026. Verfasst von Liheng Shen und 2 weiteren Autor:innen. Beteiligte Einrichtung: Department of Information Science and Technology, Graduate School of Information Science and Electrical Engineering (ISEE), Kyushu University, Ito Ca…. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 66 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Japan Science and Technology Agency (JST).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Video Anomaly Detection (VAD) is commonly formulated as a one-class classification task. Global motion, with temporal variations across most pixels within an object-centric region, e.g., walking, is typically regular, whereas localized motion, e.g., waving, can be ambiguous. Decoupled spatial and temporal jigsaw puzzles (DSTJiP) is a self-supervised method that learns discriminative representations by predicting the original order of spatially and temporally shuffled patches. However, DSTJiP’s uniform sampling and equal weighting do not assign stronger supervision to global-motion examples within the temporal objective. Consequently, the temporal supervision allocated to global-motion examples may become insufficient across training-data regimes with varying proportions of these examples, deteriorating VAD performance. Nevertheless, excessively strengthening such supervision also degrades performance. To address these issues, we propose spatial and extra temporal jigsaw puzzles (SETJiP) with two RGB-only training schemes that provide stronger and more conservative temporal supervision for global-motion examples, respectively. One scheme strengthens temporal supervision on these examples via additional temporal jigsaw puzzles. The other does so more conservatively by upweighting their temporal jigsaw puzzles. Experiments on four VAD benchmarks show that both schemes improve on DSTJiP and remain highly competitive with state-of-the-art methods.

Schlagwörter

Stichwörter der Autor:innen

self-supervised learningspatio-temporal representation learningtemporal jigsaw puzzlesvideo anomaly detection

Autor:innen und Institutionen

  • Department of Information Science and Technology, Graduate School of Information Science and Electrical Engineering (ISEE), Kyushu University, Ito Campus, 744 Motooka, Nishi-ku, Fukuoka 819-0395, Japan.
  • Department of Informatics, Faculty of Information Science and Electrical Engineering (ISEE), Kyushu University, Ito Campus, 744 Motooka, Nishi-ku, Fukuoka 819-0395, Japan.

ORCID-Profile: Liheng Shen, Tetsu Matsukawa, Einoshin Suzuki

Förderung

  • Japan Science and Technology Agency (JST)

Zitieren

APA

Shen, L., Matsukawa, T., & Suzuki, E. (2026). SETJiP: Spatial and Extra Temporal Jigsaw Puzzles for Video Anomaly Detection. Sensors, 26(9), 2889. https://doi.org/10.3390/s26092889

Vancouver

Shen L, Matsukawa T, Suzuki E. SETJiP: Spatial and Extra Temporal Jigsaw Puzzles for Video Anomaly Detection. Sensors. 2026;26(9):2889. doi:10.3390/s26092889

BibTeX
@article{Shen2026,
  author = {Shen, Liheng and Matsukawa, Tetsu and Suzuki, Einoshin},
  title = {SETJiP: Spatial and Extra Temporal Jigsaw Puzzles for Video Anomaly Detection},
  journal = {Sensors},
  year = {2026},
  volume = {26},
  number = {9},
  pages = {2889},
  doi = {10.3390/s26092889}
}
Kostenlos starten →