O-VAD: Industrial Video Anomaly Detection through Object-Centric Tracking and Reasoning

In 30 Sekunden

Was passiert ist

O-VAD detects anomalies in industrial video by tracking object states over time and reasoning about their transformations without requiring domain-specific retraining.

Warum es zählt

Manufacturing and quality control engineers need to identify defects or process violations in complex assembly or production environments with strict physical constraints.

Achtung

The paper is newly submitted and acceptance to ECCV 2026 is noted, but real-world deployment performance on proprietary production lines remains unvalidated.