Omni-AD：面向工业异常检测的大规模通用基准
Omni-AD: A Large-scale and Versatile Benchmark for Industrial Anomaly Detection
为工业异常检测提供了一个规模更大、类别更多且同时支持无监督协议与 MLLM 协议（VQA 与视觉定位）的评测平台，可直接用于检验现有 IAD 方法与多模态大模型在细粒度缺陷判别上的真实能力上限。
原文依据 The benchmark supports both conventional unsupervised and emerging MLLM-based IAD evaluation protocols.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论工业异常检测（IAD）已受到广泛关注并取得快速发展。然而，该领域的进步受到两个关键问题的制约：现有基准的性能已趋于饱和，难以对不同 IAD 方法进行有区分度的评估；以及缺乏用于评估近期多模态大语言模型（MLLM）在异常检测中表现的专用基准。为此，我们提出 Omni-AD，一个综合性 IAD 基准，具有以下特点：i）大规模：该数据集包含约 35K 张图像（为 MVTec AD 的 6 倍），涵盖 16 个工业领域的 150 个产品类别（为 MVTec AD 的 10 倍），与现有数据集相比提供了前所未有的类别多样性和更大的数据规模。ii）通用性：该基准同时支持传统的无监督评估协议和新兴的基于 MLLM 的 IAD 评估协议。后者通过定义三个难度递进的子任务实现，其中两个为视觉问答（VQA）形式，一个为视觉定位形式。iii）挑战性：对最优方法的大量实验结果表明，Omni-AD 基准比现有基准更具挑战性，可推动 IAD 领域的未来发展。
英文原文摘要 1252 字符
Industrial Anomaly Detection (IAD) has attracted significant attention and witnessed rapid development. However, the advancement in this field is hindered by two key issues: the performance saturation of existing benchmarks, limiting discriminative evaluation of different IAD methods, and the absence of benchmarks tailored to assess recent multi-modal large language models (MLLMs) in anomaly detection. To this end, we present Omni-AD, a comprehensive IAD benchmark featuring: i) Large scale: The dataset consists of approximately 35K images (6xlarger than MVTec AD) with 150 product categories (10xlarger than MVTec AD) spanning 16 industrial sectors, delivering unprecedented category diversity and greater dataset scale compared with existing datasets. ii) Versatility: The benchmark supports both conventional unsupervised and emerging MLLM-based IAD evaluation protocols. The latter is achieved by defining three subtasks of progressive difficulty, with two structured as visual question answering (VQA) and one as visual grounding. iii) Challenge: Extensive experimental results of state-of-the-art methods reveal that the Omni-AD benchmark is more challenging than existing benchmarks, which can drive the future development of the IAD field.