arXiv 2605.206822026-05-20cs.CV

IndusAgent: 通过智能工具增强开放词汇工业异常检测

IndusAgent: Reinforcing Open-Vocabulary Industrial Anomaly Detection with Agentic Tools

State Key Lab of Processors, Institute of Computing Technology, CAS（处理器国家重点实验室，计算技术研究所，中国科学院）；Santa Clara University（圣克拉拉大学）；LongCat Team（LongCat团队）；Independent Researcher（独立研究者）；New York University（纽约大学）；Sun Yat-sen University（孙中山大学）；Nanyang Technological University（南洋理工大学）；Stanford University（斯坦福大学）；University of Chinese Academy of Sciences, Beijing, China（中国科学院大学，北京，中国）

Rongbin Tan, Fangfang Lin, Zhenlong Yuan, Min Qiu, Kejin Cui, Mengmeng Wang, Yi Wang, Zijian Song, Zhiyuan Wang, Jiyuan Wang, Yue Wang, Shuhan Song§, Huawei Cao

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出IndusAgent框架，通过整合视觉观测、高分辨率局部片段和专家正常性先验，提升开放词汇工业异常检测的零样本性能，验证了方法的鲁棒性和泛化能力。

摘要：

多模态大语言模型（MLLMs）在连接视觉感知和文本推理方面表现出色，能够跨多样化的工业场景实现零样本理解。然而，其在开放词汇工业异常检测（IAD）中的性能常受限于领域不匹配的推理和幻觉的结构推断。为了解决这些挑战，我们提出了IndusAgent，一种工具增强的智能框架用于开放词汇IAD。具体而言，我们首先构建了Indus-CoT，一个整合了全局视觉观测、高分辨率局部片段和专家正常性先验的结构化数据集，为在严格工业检查轨迹上微调模型提供监督。在此基础上，IndusAgent动态协调一组外部工具，包括动态区域裁剪、高频特征增强和先验检索，从而使代理能够主动解决视觉歧义并分离细微异常。此外，我们引入了一个门控强化学习目标，联合优化异常分类、定位准确性、异常类型推理和高效的工具使用，确保工具调用仅在有益时发生。在五个工业异常基准测试上（包括MVTec-AD、VisA、MPDD、DTD和SDD）的广泛评估表明，IndusAgent在所有现有方法中实现了最先进的零样本性能，验证了我们的鲁棒性和泛化能力。

英文摘要：

Multimodal large language models (MLLMs) have shown remarkable capability in bridging visual perception and textual reasoning, enabling zero-shot understanding across diverse industrial scenarios. However, their performance in open-vocabulary industrial anomaly detection (IAD) is often limited by domain-misaligned reasoning and hallucinated structural inferences. To address these challenges, we propose \textbf{IndusAgent}, a tool-augmented agentic framework for open-vocabulary IAD. Specifically, we first construct \textbf{Indus-CoT}, a structured dataset that integrates global visual observations, high-resolution local patches, and expert normalcy priors, providing supervision for fine-tuning the model on rigorous industrial inspection trajectories. Building on this, IndusAgent dynamically orchestrates a set of external tools, including dynamic region cropping, high-frequency feature enhancement, and prior retrieval, thus enabling the agent to actively resolve visual ambiguities and disentangle subtle anomalies. Furthermore, we introduce a gated reinforcement learning objective that jointly optimizes anomaly classification, localization accuracy, anomaly type reasoning, and efficient tool usage, ensuring that tool invocation occurs only when beneficial. Extensive evaluations on five industrial anomaly benchmarks, including MVTec-AD, VisA, MPDD, DTD, and SDD, demonstrate that IndusAgent achieves state-of-the-art zero-shot performance among all existing methods, validating our robustness and generalization capacity.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。