MoECLIP：面向零样本异常检测的图像块专用专家
MoECLIP: Patch-Specialized Experts for Zero-shot Anomaly Detection
把 MoE 路由下沉到图像块级别、每个图像块动态选择专属 LoRA 专家，是给 CLIP 类视觉语言模型做任务特化的可复用微调范式；FOFS 正交特征分离与 simplex ETF 损失抑制专家冗余的思路，可直接迁移到其他需要多专家分工的密集预测任务。
原文依据 dynamically routing each image patch to a specialized Low-Rank Adaptation (LoRA) expert based on its unique characteristics
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论CLIP 模型出色的泛化能力推动了零样本异常检测（ZSAD）在未见类别上的近期进展。ZSAD 的核心挑战在于：在保留 CLIP 强大泛化能力的同时，使模型专门适配异常检测任务。现有方法在解决这一挑战时都存在一个根本局限，即采用与图像块无关的设计，将所有图像块作为一个整体统一处理，而不考虑它们各自的独特特性。为解决这一局限，我们提出 MoECLIP，一种面向 ZSAD 任务的混合专家（MoE）架构，它依据每个图像块的独特特性将其动态路由到专门的低秩适配（LoRA）专家，从而实现图像块级别的适配。此外，为防止 LoRA 专家之间出现功能冗余，我们引入（1）冻结正交特征分离（FOFS），通过正交方式分离输入特征空间，迫使各专家关注不同的信息；（2）simplex 等角紧框架（ETF）损失，约束专家输出形成最大等角表示。在涵盖工业与医学领域的 14 个基准数据集上的全面实验结果表明，MoECLIP 优于现有最优方法。代码见 https://github.com/CoCoRessa/MoECLIP。
英文原文摘要 1380 字符
The CLIP model's outstanding generalization has driven recent success in Zero-Shot Anomaly Detection (ZSAD) for detecting anomalies in unseen categories. The core challenge in ZSAD is to specialize the model for anomaly detection tasks while preserving CLIP's powerful generalization capability. Existing approaches attempting to solve this challenge share the fundamental limitation of a patch-agnostic design that processes all patches monolithically without regard for their unique characteristics. To address this limitation, we propose MoECLIP, a Mixture-of-Experts (MoE) architecture for the ZSAD task, which achieves patch-level adaptation by dynamically routing each image patch to a specialized Low-Rank Adaptation (LoRA) expert based on its unique characteristics. Furthermore, to prevent functional redundancy among the LoRA experts, we introduce (1) Frozen Orthogonal Feature Separation (FOFS), which orthogonally separates the input feature space to force experts to focus on distinct information, and (2) a simplex equiangular tight frame (ETF) loss to regulate the expert outputs to form maximally equiangular representations. Comprehensive experimental results across 14 benchmark datasets spanning industrial and medical domains demonstrate that MoECLIP outperforms existing state-of-the-art methods. The code is available at https://github.com/CoCoRessa/MoECLIP.