arXiv:2512.06840v1 Announce Type: new 
Abstract: Video anomaly detection (VAD) has long been studied as a crucial problem in public security and crime prevention. In recent years, weakly-supervised VAD (WVAD) have attracted considerable attention due to their easy annotation process and promising research results. While existing WVAD methods tackle mainly on static datasets, the possibility that the domain of data can vary has been neglected. To adapt such domain-shift, the continual learning (CL) perspective is required because otherwise additional training only with new coming data could easily cause performance degradation for previous data, i.e., forgetting. Therefore, we propose a brand-new approach, called Continual Anomaly Detection with Ensembles (CADE) that is the first work combining CL and WVAD viewpoints. Specifically, CADE uses the Dual-Generator(DG) to address data imbalance and label uncertainty in WVAD. We also found that forgetting exacerbates the "incompleteness'' where the model becomes biased towards certain anomaly modes, leading to missed detections of various anomalies. To address this, we propose to ensemble Multi-Discriminator (MD) that capture missed anomalies in past scenes due to forgetting, using multiple models. Extensive experiments show that CADE significantly outperforms existing VAD methods on the common multi-scene VAD datasets, such as ShanghaiTech and Charlotte Anomaly datasets.

تم اقتراح نهج جديد يسمى الكشف المستمر عن الشذوذ باستخدام الفرق (CADE) لتحسين الكشف عن الشذوذ في الفيديو تحت إشراف ضعيف (WVAD) من خلال دمج تقنيات التعلم المستمر (CL). يتناول هذا الأسلوب التحديات مثل عدم توازن البيانات وعدم اليقين في التسميات، والتي تنتشر في الأساليب الحالية لـ WVAD التي تركز بشكل أساسي على مجموعات البيانات الثابتة.

Se ha propuesto un nuevo enfoque llamado Detección Continua de Anomalías con Conjuntos (CADE) para mejorar la detección de anomalías en video débilmente supervisada (WVAD) mediante la integración de técnicas de aprendizaje continuo (CL). Este método aborda desafíos como el desequilibrio de datos y la incertidumbre de etiquetas, que son comunes en los métodos WVAD existentes que se centran principalmente en conjuntos de datos estáticos.

Une nouvelle approche appelée Détection Anormale Continue avec Ensembles (CADE) a été proposée pour améliorer la détection d'anomalies vidéo faiblement supervisée (WVAD) en intégrant des techniques d'apprentissage continu (CL). Cette méthode s'attaque à des défis tels que le déséquilibre des données et l'incertitude des étiquettes, qui sont répandus dans les méthodes WVAD existantes qui se concentrent principalement sur des ensembles de données statiques.

A new approach called Continual Anomaly Detection with Ensembles (CADE) has been proposed to enhance weakly-supervised video anomaly detection (WVAD) by integrating continual learning (CL) techniques. This method addresses challenges such as data imbalance and label uncertainty, which are prevalent in existing WVAD methods that primarily focus on static datasets.

CADE: Continual Weakly-supervised Video Anomaly Detection with Ensembles

Was this article worth reading? Share it

LucidQuery AI

Videotok

Videolulu

Ready to build your own newsroom?