EdgeZSAD：边缘设备上的实用零样本异常检测
EdgeZSAD: Practical Zero-Shot Anomaly Detection on Edge Devices
AI总结：
针对边缘部署约束，提出基于TinyViT-21M-512骨干、非对称全局-局部读出（EdgeGLR）和可复现源训练方案（Real-IAD-DR）的紧凑零样本异常检测系统，在多个工业基准上达到高精度且可直接部署。
摘要：
工业检测需要零样本异常检测（ZSAD），该检测在边缘部署约束下仍然有效。最近的方法通常依赖ViT-L基础骨干（约3亿参数），这超出了典型嵌入式硬件的内存和算子预算。我们通过EdgeZSAD研究这一场景，这是一个紧凑的参考系统，围绕TinyViT-21M-512骨干、非对称全局-局部读出（EdgeGLR）和可复现的源端训练方案（Real-IAD-DR）构建。我们在源训练、目标未见协议下训练单个检查点，并在六个工业基准上评估。在三次独立运行中，所得模型在MVTec-AD上平均图像AUROC达到91.6，在VisA上达到88.2，同时可直接部署在Jetson Orin Nano Super（TensorRT FP16）和RB5 Gen2（QNN GPU FP16）上。在六个设备重新评分的基准中，图像AUROC漂移保持在0.2点以下，表明导出的图在评估的部署设置中保留了主机端的排序行为。
英文摘要：
Industrial inspection needs zero-shot anomaly detection (ZSAD) that remains useful under edge deployment constraints. Recent methods often rely on ViT-L foundation backbones (~300M parameters), which exceed the memory and operator budget of typical embedded hardware. We study this regime through EdgeZSAD, a compact reference system built around a TinyViT-21M-512 backbone, an asymmetric global-local readout (EdgeGLR), and a reproducible source-side training recipe (Real-IAD-DR). We train a single checkpoint in a source-trained, target-unseen protocol and evaluate it across six industrial benchmarks. Across three independent runs, the resulting model reaches an average image AUROC of 91.6 on MVTec-AD and 88.2 on VisA, while remaining directly deployable on Jetson Orin Nano Super (TensorRT FP16) and RB5 Gen2 (QNN GPU FP16). Across the six device-rescored benchmarks, image-AUROC drift stays below 0.2 points, indicating that the exported graph preserves host-side ranking behavior in the evaluated deployment setting.
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。
