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

2026-07-20

AI Summary

Industrial video anomaly detection is challenged by complex object dynamics, strict physical constraints, and procedural rules. This work proposes a training-free agent framework that, for the first time, incorporates a human inspector–inspired reasoning mechanism for object state evolution into this task. Leveraging a vision-language model, the approach performs object-centric spatiotemporal trajectory tracking and zero-shot state reasoning without requiring domain-specific knowledge or retraining on normal samples. Evaluated on three industrial video anomaly detection benchmarks, the method outperforms existing vision-language models, agent-based approaches, and fine-tuned baselines, while also generating interpretable reports that detail the type and progression of detected anomalies.

Abstract

Industrial Video Anomaly Detection (IVAD) aims to identify anomalous objects and events in an industrial process, which is crucial for modern manufacturing and quality control systems. Existing VLM-based anomaly reasoning methods are capable of detecting open-ended anomalies in general domains. However, their performance declines in industrial settings characterized by intricate object transformations, strict physics, and procedural constraints. To tackle the complexity of such interaction-intensive detection, we introduce a training-free agentic framework for anomaly detection free of domain-specific knowledge, emphasizing object state evolution like humans inspectors. It is designed to track spatial-temporal dynamics and underlying transformations of detected objects over time, and then reason over the object-wise temporal state trajectories to identify abnormal objects in grounded frames. Our method overcomes limitations of prior approaches that rely on retraining on normal clips or injecting domain knowledge as context for test-time inference. Extensive experiments on three IVAD datasets demonstrate that our method outperforms frontier VLMs, agentic frameworks, and traditional VAD methods fine-tuned on the respective datasets, while providing interpretable reports over anomaly processes and types.

Problem

• Industrial Video Anomaly Detection
• Object-Centric Tracking
• Anomaly Reasoning
• Temporal State Trajectories
• Complex Object Transformations

Innovation

• Object-Centric Tracking
• Training-Free Anomaly Detection
• Temporal State Reasoning
• Industrial Video Anomaly Detection
• Interpretable AI