EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

X-VAD: Video Anomaly Detection with Kinematic Explanations

Melisa Gözet iD, Asım Egemen Yılmaz iD, Mehmet Karaköse iD

Sensors · 24.10.2026 · 26 (21) · 6731

DOI: 10.3390/s26216731 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 24.10.2026. Verfasst von Melisa Gözet und 2 weiteren Autor:innen. Beteiligte Einrichtung: Department of Artificial Intelligence, Graduate School of Natural and Applied Sciences, Ankara University, 06110 Ankara, Türkiye. Die Arbeit nennt 43 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Türkiye Bilimsel ve Teknolojik Araştırma Kurumu.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Anomaly scores alone do not describe unusual motion. X-VAD links scoring and reporting through 13 cell-level kinematic descriptors from SEA-RAFT optical flow. They describe speed, direction, flow deformation, and temporal change. Each vector is compared with normal training observations from the same cell. The four highest calibrated cell scores are averaged to obtain the frame score. No gradient-based training of an anomaly scorer is required. Rule-based alarm reports identify the region, descriptor deviations, and nearest normal reference. YOLO-World adds object context without changing the score. Micro-averaged area under the receiver operating characteristic curve (AUROC) reaches 85.78%, 84.42%, and 75.43% on UCSD Ped2, CUHK Avenue, and IITB-Corridor, respectively. At the 0.85 threshold, all stored test scores yield true-positive rates of 87.69%, 90.78%, and 62.73%. Corresponding false-positive rates are 41.46%, 46.32%, and 26.53%. Runtime measurements use an NVIDIA A100-SXM4-40GB at batch size 1. Mean score-availability times are 50.0, 50.6, and 82.6 ms per frame pair, excluding frame reading. Report generation, including YOLO-World inference, adds 16.3–18.3 ms per alarm frame on average. Visualization and notification delivery are excluded. The IITB-Corridor AUROC approaches VideoPatchCore’s 76.4%.X-VAD provides inspectable motion reports, although its AUROCs trail SETJiP on Ped2 (99.3%) and AI-VAD on Avenue (93.7%).

Schlagwörter

Stichwörter der Autor:innen

explainable artificial intelligencemotion analysisoptical flowvideo anomaly detectionvideo surveillance

Autor:innen und Institutionen

  • Department of Artificial Intelligence, Graduate School of Natural and Applied Sciences, Ankara University, 06110 Ankara, Türkiye.
  • Department of Artificial Intelligence Engineering, Faculty of Engineering, OSTİM Technical University, 06374 Ankara, Türkiye.
  • Department of Electrical and Electronics Engineering, Faculty of Engineering, Ankara University, 06830 Ankara, Türkiye.
  • Department of Computer Engineering, Fırat University, 23119 Elazığ, Türkiye.

ORCID-Profile: Melisa Gözet, Asım Egemen Yılmaz, Mehmet Karaköse

Förderung

  • Türkiye Bilimsel ve Teknolojik Araştırma Kurumu

Zitieren

APA

Gözet, M., Yılmaz, A. E., & Karaköse, M. (2026). X-VAD: Video Anomaly Detection with Kinematic Explanations. Sensors, 26(21), 6731. https://doi.org/10.3390/s26216731

Vancouver

Gözet M, Yılmaz AE, Karaköse M. X-VAD: Video Anomaly Detection with Kinematic Explanations. Sensors. 2026;26(21):6731. doi:10.3390/s26216731

BibTeX
@article{Gzet2026,
  author = {Gözet, Melisa and Yılmaz, Asım Egemen and Karaköse, Mehmet},
  title = {X-VAD: Video Anomaly Detection with Kinematic Explanations},
  journal = {Sensors},
  year = {2026},
  volume = {26},
  number = {21},
  pages = {6731},
  doi = {10.3390/s26216731}
}
Kostenlos starten →