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

O-VAD is a training-free framework for industrial video anomaly detection that tracks objects over time and reasons about state changes.

By focusing on object-centric tracking and reasoning, it avoids relying on domain-specific knowledge or retraining on normal clips.

The approach aims to handle complex object interactions better than existing VLM-based methods in industrial settings.

It also produces interpretable outputs that can support quality control and inspection.