arXiv 2504.067402025-08-06cs.CV

MultiADS：零样本学习中多类型异常检测与分割的缺陷感知监督

MultiADS: Defect-aware Supervision for Multi-type Anomaly Detection and Segmentation in Zero-Shot Learning

University of Vienna（维也纳大学）；Bosch Corporate Research, Robert Bosch GmbH（博世公司博世企业研究部）；University of Stuttgart（斯图加特大学）；University of Southampton（南安普顿大学）

Ylli Sadikaj, Hongkuan Zhou, Lavdim Halilaj, Stefan Schmid, Steffen Staab, Claudia Plant

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

提出MultiADS，首个零样本多类型异常检测与分割方法，利用CLIP和线性层对齐视觉文本特征，生成特定缺陷掩码并区分多种缺陷类型，在五个数据集上超越现有方法。

摘要：

在工业应用中，精确的光学检测对于最大限度地降低废品率和相关成本至关重要。除了仅仅检测产品是否存在异常外，了解缺陷的具体类型（如弯曲、切割或划痕）也至关重要。识别“确切”缺陷类型的能力使得现代生产线能够对异常进行自动化处理。当前方法仅限于检测产品是否有缺陷，而不提供关于缺陷类型的任何信息，更不用说检测和识别多种缺陷了。我们提出了MultiADS，一种零样本学习方法，能够执行多类型异常检测与分割。MultiADS的架构包括CLIP和额外的线性层，以在联合特征空间中对齐视觉和文本表示。据我们所知，我们的方法是首个在零样本学习中执行多类型异常分割任务的方法。与其他基线方法不同，我们的方法i)为每种不同的缺陷类型生成特定的异常掩码，ii)学习区分缺陷类型，iii)同时识别异常产品中存在的多种缺陷类型。此外，我们的方法在五个常用数据集（MVTec-AD、Visa、MPDD、MAD和Real-IAD）上的图像级和像素级异常检测与分割任务中，优于零样本/少样本学习的SoTA方法。

英文摘要：

Precise optical inspection in industrial applications is crucial for minimizing scrap rates and reducing the associated costs. Besides merely detecting if a product is anomalous or not, it is crucial to know the distinct type of defect, such as a bent, cut, or scratch. The ability to recognize the "exact" defect type enables automated treatments of the anomalies in modern production lines. Current methods are limited to solely detecting whether a product is defective or not without providing any insights on the defect type, nevertheless detecting and identifying multiple defects. We propose MultiADS, a zero-shot learning approach, able to perform Multi-type Anomaly Detection and Segmentation. The architecture of MultiADS comprises CLIP and extra linear layers to align the visual- and textual representation in a joint feature space. To the best of our knowledge, our proposal, is the first approach to perform a multi-type anomaly segmentation task in zero-shot learning. Contrary to the other baselines, our approach i) generates specific anomaly masks for each distinct defect type, ii) learns to distinguish defect types, and iii) simultaneously identifies multiple defect types present in an anomalous product. Additionally, our approach outperforms zero/few-shot learning SoTA methods on image-level and pixel-level anomaly detection and segmentation tasks on five commonly used datasets: MVTec-AD, Visa, MPDD, MAD and Real-IAD.

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。