VT-3DAD: Cross-Category 3D Anomaly Detection via Visual-Text Normal Space Alignment

2026-06-02

AI Summary

This work addresses the challenge of cross-category 3D point cloud anomaly detection using only a few normal samples by proposing a training-free framework that, for the first time, introduces vision–text normal space alignment to this task. The method extracts visual features from multi-view depth maps using a frozen CLIP model and constructs semantic normal anchors via depth-aware and 3D-aware textual prompts. Anomaly scores are computed by fusing visual and semantic deviations. Evaluated on ShapeNetPart, the approach achieves an average single-sample AUC-ROC of 94.80%, outperforming a purely visual baseline by 2.31% and reducing the standard deviation from 5.64 to 3.41. These results demonstrate the effectiveness of the proposed method in jointly modeling geometric and semantic normality, as well as its strong cross-category generalization capability.

Abstract

Few-shot cross-category 3D anomaly detection aims to determine whether an unknown point cloud belongs to a target normal category using only a few normal references. Existing training-based methods usually require category-wise optimization, while recent training-free methods based on multi-view CLIP visual features mainly rely on visual similarity and may be confused by geometrically similar categories. In this paper, we propose VT-3DAD, a training-free framework for cross-category 3D anomaly detection via Visual-Text Normal Space Alignment. Given few-shot normal references and a test point cloud, VT-3DAD first generates realistic multi-view depth maps and extracts view-wise features using a frozen CLIP visual encoder. The visual branch measures reference-test deviation in the multi-view feature space. In parallel, depth-aware and 3D-aware prompts are encoded by the frozen CLIP text encoder to construct textual normal anchors, which provide semantic normality constraints for the target category. The final anomaly score is obtained by fusing visual deviation from normal references and semantic deviation from the textual normal space. Experiments on the ShapeNetPart dataset demonstrate that VT-3DAD achieves state-of-the-art performance. In particular, VT-3DAD improves the one-shot average AUC-ROC from 92.49% to 94.80% compared with the visual-only baseline, while also reducing the average standard deviation from 5.64 to 3.41.

Problem

• 3D anomaly detection
• cross-category
• few-shot
• point cloud
• normality assessment

Innovation

• 3D anomaly detection
• few-shot learning
• CLIP
• visual-text alignment
• training-free