CEAVAD: Contrastive Event Adjudication for Training-Free Video Anomaly Detection

Forum topic · 小凯 · 2026-08-11

Summary

This paper introduces CEAVAD, a training-free framework for video anomaly detection (VAD) that localizes abnormal events in video. The authors argue that existing training-free VAD methods leverage pretrained models for semantic understanding but lack a clear decision criterion for abnormality, since richer descriptions capture hazard resemblance without resolving anomaly status. CEAVAD reframes inference from isolated anomaly concepts to falsifiable event hypotheses, establishing an explanatory boundary at inference time. It first constructs hazard-benign event contrasts from public safety knowledge, pairing each hazard mechanism with a generic normal description and a mechanism-specific benign counterpart. It then determines whether a target interval better supports the hazard interpretation or its benign competitor, producing a revisable contrastive boundary proposal. Finally, the framework adjudicates between competing explanations to verify whether the hazard hypothesis holds against video evidence, enabling temporally localized anomaly detection and evidence-based explanation. Experiments on three VAD benchmarks show CEAVAD achieves state-of-the-art performance in the training-free paradigm, eliminating the need for target-domain annotations.

Paper Overview

Field: Computer Vision (CV)
Authors: Wenti Yin, Xiang Wang, Huaxin Zhang
Published: 2026-08-11
arXiv: 2508.03800

Summary

Video anomaly detection (VAD) aims to identify and temporally localize abnormal events in videos. Supervised methods learn anomaly decision boundaries from target-domain annotations but require substantial in-domain data. Existing training-free methods leverage the rich semantic knowledge and reasoning capabilities of pretrained models to interpret visual content. However, these capabilities do not directly define an anomaly decision criterion: richer anomaly descriptions better capture hazard resemblance without resolving abnormality.

To address this, the authors propose Contrastive Event Adjudication for training-free Video Anomaly Detection (CEAVAD), which shifts the unit of inference from isolated anomaly concepts to falsifiable event hypotheses and establishes an inference-time explanatory boundary through the interaction between competing explanations and video evidence.

Key Points

Problem: Training-free VAD methods describe hazards well but lack a principled criterion for distinguishing abnormal from normal events.
Approach: CEAVAD reframes anomaly detection as hypothesis adjudication between hazard and benign event interpretations of the same visual evidence.
Step 1 – Hazard-benign contrast construction: Uses public safety knowledge to pair each hazard mechanism with (a) a generic normal description and (b) a mechanism-specific benign counterpart.
Step 2 – Contrastive boundary proposal: Determines whether a target interval better supports the hazard interpretation or its benign competitor, producing a revisable boundary proposal.
Step 3 – Adjudication: Competing explanations are evaluated against video evidence to verify whether the hazard hypothesis is falsified, enabling both temporally localized anomaly detection and evidence-based explanation.
Results: Experiments on three widely used VAD benchmarks show CEAVAD achieves state-of-the-art performance in the training-free paradigm, without requiring target-domain annotations.

Source

arXiv link: https://arxiv.org/abs/2508.03800
Auto-collected: 2026-08-12

CEAVAD: Contrastive Event Adjudication for Training-Free Video Anomaly Detection

Summary

Paper Overview

Summary

Key Points

Source

Tags