Referring Industrial Anomaly Segmentation

TL;DR

The paper introduces RIAS, a language-guided anomaly segmentation method that produces precise masks without manual thresholds, supported by a new dataset and a novel transformer-based benchmark for industrial anomaly detection.

Contribution

It presents RIAS, a novel language-guided approach for industrial anomaly segmentation, along with the MVTec-Ref dataset and the DQFormer benchmark with LMA for improved multi-scale detection.

Findings

RIAS achieves precise, threshold-free anomaly segmentation.

The MVTec-Ref dataset includes diverse referring expressions and small anomalies.

DQFormer with LMA enhances multi-scale anomaly detection.

Abstract

Industrial Anomaly Detection (IAD) is vital for manufacturing, yet traditional methods face significant challenges: unsupervised approaches yield rough localizations requiring manual thresholds, while supervised methods overfit due to scarce, imbalanced data. Both suffer from the "One Anomaly Class, One Model" limitation. To address this, we propose Referring Industrial Anomaly Segmentation (RIAS), a paradigm leveraging language to guide detection. RIAS generates precise masks from text descriptions without manual thresholds and uses universal prompts to detect diverse anomalies with a single model. We introduce the MVTec-Ref dataset to support this, designed with diverse referring expressions and focusing on anomaly patterns, notably with 95% small anomalies. We also propose the Dual Query Token with Mask Group Transformer (DQFormer) benchmark, enhanced by Language-Gated Multi-Level Aggregation (LMA) to improve multi-scale segmentation. Unlike traditional methods using redundant queries, DQFormer employs only "Anomaly" and "Background" tokens for efficient visual-textual integration. Experiments demonstrate RIAS's effectiveness in advancing IAD toward open-set capabilities. Code: https://github.com/swagger-coder/RIAS-MVTec-Ref.