发表机构
Actual Reality Technologies（Actual Reality Technologies）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出利用地面真值缺陷掩码作为空间监督信号，通过注意力对齐损失引导CNN特征图聚焦缺陷区域，结合DDPM增强，在MVTec-AD上显著提升定位性能，且不影响分类精度。
AI 中文摘要
工业检测数据集中的地面真值缺陷掩码通常保留用于评估。本文将其重新用作分类网络训练期间的空间监督信号，教导模型不仅预测什么，而且关注哪里。该方法添加了一个基于激活的注意力对齐损失，将卷积特征图引导至缺陷区域，采用混合监督公式，该公式也适用于没有掩码的样本，例如扩散生成的图像。与DDPM增强相结合，合成图像贡献数量，而掩码贡献空间精度。我们在MVTec-AD瓶子基准上评估了85个模型（四个CNN骨干网络，在五个种子下进行2x2数据/训练因子设计，加上Swin-V2-T变压器基线），定位在排除分类器梯度更新的保留缺陷图像上测量。主要发现：（1）注意力引导训练改善了基于激活的定位（Pixel-AUROC），对于EfficientNetB0（增强）提高了+18.0%（p=0.005，Cohen's d=2.6），对于ResNet50提高了+18.7%（p=0.008），在八个CNN设置中有四个显著（未校正多重比较），分类性能无显著变化；（2）对于EfficientNetB0，数据x训练模式交互显著（p=0.002），与超加性效应一致（组合+13.6% vs 单独效应总和+1.6%）；（3）空间表示较弱的架构受益最多，而ConvNeXt-T没有效果，显然是因为其深度卷积激活产生空间信息不足的通道均值图；（4）无监督PatchCore仍然是最强的定位器（Pixel-AUROC=0.983），为监督增益提供了背景。这些结果表明，现有的评估掩码可以作为实用的训练信号，可测量且可重复地改善缺陷分类器的关注位置。
英文摘要
Ground-truth defect masks in industrial inspection datasets are typically reserved for evaluation. This paper repurposes them as spatial supervision signals during training of classification networks, teaching a model not just what to predict but where to look. The method adds an activation-based attention alignment loss that steers convolutional feature maps toward defect regions, in a mixed-supervision formulation that also accommodates samples without masks, such as diffusion-generated images. Combined with DDPM augmentation, synthetic images contribute quantity while masks contribute spatial precision. We evaluate 85 models (four CNN backbones under a 2x2 data/training factorial over five seeds, plus a Swin-V2-T transformer baseline) on the MVTec-AD bottle benchmark, with localization measured on held-out defect images excluded from classifier gradient updates. Main findings: (1) attention-guided training improves activation-based localization (Pixel-AUROC) by +18.0% for EfficientNetB0 with augmentation (p=0.005, Cohen's d=2.6) and +18.7% for ResNet50 (p=0.008), significant in four of eight CNN settings (uncorrected for multiple comparisons) with no significant change in classification; (2) for EfficientNetB0 a data x training-mode interaction is significant (p=0.002), consistent with a super-additive effect (+13.6% combined vs +1.6% summed individual effects); (3) architectures with weaker spatial representations benefit most, whereas ConvNeXt-T shows no effect, apparently because its depthwise-convolution activations yield spatially uninformative channel-mean maps; (4) unsupervised PatchCore remains the strongest localizer (Pixel-AUROC=0.983), contextualizing the supervised gains. These results show that existing evaluation masks can act as practical training signals that measurably and reproducibly improve where defect classifiers attend.
Comments20 pages, 5 figures, 5 tables. Code: https://github.com/Actual-Reality/Glass-Defect-Detection-Attention-Supervision
