IndusAgent: Reinforcing Open-Vocabulary Industrial Anomaly Detection with Agentic Tools
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910239760056320 |
|---|---|
| author | Tan, Rongbin Lin, Fangfang Yuan, Zhenlong Qiu, Min Cui, Kejin Wang, Mengmeng Wang, Yi Song, Zijian Wang, Zhiyuan Wang, Jiyuan Wang, Yue Song§, Shuhan Cao, Huawei |
| author_facet | Tan, Rongbin Lin, Fangfang Yuan, Zhenlong Qiu, Min Cui, Kejin Wang, Mengmeng Wang, Yi Song, Zijian Wang, Zhiyuan Wang, Jiyuan Wang, Yue Song§, Shuhan Cao, Huawei |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_20682 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | IndusAgent: Reinforcing Open-Vocabulary Industrial Anomaly Detection with Agentic Tools Tan, Rongbin Lin, Fangfang Yuan, Zhenlong Qiu, Min Cui, Kejin Wang, Mengmeng Wang, Yi Song, Zijian Wang, Zhiyuan Wang, Jiyuan Wang, Yue Song§, Shuhan Cao, Huawei Computer Vision and Pattern Recognition 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. |
| title | IndusAgent: Reinforcing Open-Vocabulary Industrial Anomaly Detection with Agentic Tools |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2605.20682 |