发表机构
Shanghai University; National University of Singapore; University of Electronic Science and Technology of China（上海大学; 新加坡国立大学; 电子科技大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
研究多模态工业异常检测问题，提出TC-MAF方法，通过固定像素级融合公式结合多模态检测器等，利用轻量级TDC项缩放辅助参与度，在MVTec-3D上取得优异检测和定位结果，消融实验揭示关键因素及校准增益。
AI 中文摘要
多模态异常检测受益于互补的RGB和3D证据，但辅助RGB重建在不同产品类别中可靠性不同，且通常没有按类别测试时的策略选择。我们提出了TC-MAF，一种基于固定像素级融合公式，结合多模态检测器、互补的Dinomaly证据和小的跨模态一致性线索的基础锚定多证据融合设计。一个轻量级的训练离散置信度（TDC）项仅使用正常训练统计来缩放辅助参与度。在MVTec-3D上，TC-MAF达到了0.979的图像级AUROC和0.990的像素级AUPRO，在比较的多模态方法中在检测和定位方面均取得了最佳平均结果。系统消融表明融合结构本身是主导因素，而TDC相对于无校准或任意校准提供了较小但可重复的校准增益。额外实验表明相同设计在合并统计变体、辅助分支和主干替换、少样本设置、缺少3D设置以及对Eyecandies的跨数据集评估下仍然有效。代码可在指定链接获取。
英文摘要
Multimodal anomaly detection benefits from complementary RGB and 3D evidence, yet auxiliary RGB reconstruction is not equally reliable across product categories and class-wise test-time policy selection is usually unavailable. We propose TC-MAF, a base-anchored multi-evidence fusion design that combines a multimodal detector, complementary Dinomaly evidence, and a small cross-modal consistency cue under one fixed pixel-level fusion formula. A lightweight training-dispersion confidence (TDC) term scales auxiliary participation using only normal training statistics. On MVTec-3D, TC-MAF reaches 0.979 image-level AUROC and 0.990 pixel-level AUPRO, achieving the best mean results on both detection and localization among the compared multimodal methods. Systematic ablations show that the fusion structure itself is the dominant factor, while TDC provides a smaller but reproducible calibration gain over no calibration or arbitrary calibration. Additional experiments show that the same design remains effective under a pooled-statistics variant, auxiliary-branch and backbone substitutions, few-shot settings, a missing-3D setting, and cross-dataset evaluation on Eyecandies. Code is available at https://anonymous.4open.science/r/TC_MAF-C3BB.
Commentsaccepted by ACM MM 2026
