面向多类别 3D 异常检测的语义解耦统一模型
A Semantically Disentangled Unified Model for Multi-category 3D Anomaly Detection
针对统一多类别 3D 异常检测中的类别间语义纠缠问题，提出基于解耦语义表示的条件重建框架，可复用于其他需要多类别统一建模的 3D 点云分析任务。
原文依据 we propose the Semantically Disentangled Unified Model for 3D Anomaly Detection, which reconstructs features conditioned on disentangled semantic representations.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论3D 异常检测旨在仅使用正常数据训练的情况下，检测并定位 3D 点云中的缺陷。尽管统一模型通过跨多个类别学习提升了可扩展性，但它常常遭受类别间纠缠（ICE）问题——不同类别的潜变量特征发生重叠，导致模型在重建过程中采用错误的语义先验，最终产生不可靠的异常分数。为解决该问题，我们提出了面向 3D 异常检测的语义解耦统一模型，该模型以解耦的语义表示作为条件进行特征重建。我们的框架包含三个关键组件：（i）由粗到细的全局标记化，用于形成实例级语义身份；（ii）类别条件对比学习，用于解耦类别语义；（iii）几何引导解码器，用于语义一致的重建。在 Real3D-AD 和 Anomaly-ShapeNet 上的大量实验表明，我们的方法在统一模型和类别特定模型上均达到了最优性能，分别将对象级 AUROC 提升了 2.8% 和 9.1%，同时增强了统一 3D 异常检测的可靠性。
英文原文摘要 1210 字符
3D anomaly detection targets the detection and localization of defects in 3D point clouds trained solely on normal data. While a unified model improves scalability by learning across multiple categories, it often suffers from Inter-Category Entanglement (ICE)--where latent features from different categories overlap, causing the model to adopt incorrect semantic priors during reconstruction and ultimately yielding unreliable anomaly scores. To address this issue, we propose the Semantically Disentangled Unified Model for 3D Anomaly Detection, which reconstructs features conditioned on disentangled semantic representations. Our framework consists of three key components: (i) Coarse-to-Fine Global Tokenization for forming instance-level semantic identity, (ii) Category-Conditioned Contrastive Learning for disentangling category semantics, and (iii) a Geometry-Guided Decoder for semantically consistent reconstruction. Extensive experiments on Real3D-AD and Anomaly-ShapeNet demonstrate that our method achieves state-of-the-art for both unified and category-specific models, improving object-level AUROC by 2.8% and 9.1%, respectively, while enhancing the reliability of unified 3D anomaly detection.