mlsanityData to 14 Oct 2026

CMDS-AD: Cross-Modal Dual-Stream Decoupling for Few-Shot Anomaly Detection

arXiv 2606.20300 ↗First posted 18 Jun 2026cs.CVcited by 0

On the map: Diffusion models → Flow matching, time series and science

Abstract (arXiv)

Few-shot anomaly detection remains challenging due to limited training data. Multi-modal anomaly detection (MAD) offers a viable solution, leveraging 3D geometric cues to enrich 2D RGB representations and compensate for this scarcity. However, existing MAD methods apply spatially uniform feature processing, conflating stable macroscopic structures with high-frequency localized defect signals, exacerbating cross-modal misalignment and inflating false-positive rates. To overcome this, we present CMDS-AD, a Cross-Modal Dual-Stream Anomaly Detection framework. A LoRA-guided diffusion model generates diverse RGB samples to mitigate extreme data scarcity. For 3D normal augmentation, we employ a pre-trained diffusion model as a normal estimator. Crucially, this estimator inherently acts as a non-linear low-pass filter, directly extracting low-frequency normal representations from RGB inputs. This establishes an auxiliary estimated stream of purely low-frequency information, anchoring robust structural templates and assisting the uncompressed real stream, containing coupled high- and low-frequency components, to precisely isolate micro-defects. A Coordinate-Aware Hierarchical Feature Mapper adaptively aligns cross-modal semantics, while a multiplicative scoring mechanism filters modality-specific noise. Under the extreme 1-shot setting, CMDS-AD achieves absolute performance gains of 5.7% (I-AUROC) and 2.0% (AUPRO) on MVTec 3D-AD, alongside 7.7% and 5.6% improvements on EyeCandies, establishing a new state-of-the-art. Code is available at https://github.com/Junhaocai27/CMDS-AD

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Flow matching, time series and science, one of 14 directions in Diffusion models. Builds on 29 of its 41 references that are in the map. First posted on arXiv 18 Jun 2026.

Builds on

29 of 41 references.

Closest ancestors · 1

Same direction, newest first

  1. 2023DiffusionAD: Norm-guided One-step Denoising Diffusion for Anomaly Detection15 Mar

Other references · 28

  1. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,380
  2. 2022Anomaly Detection via Reverse Distillation from One-Class Embeddingcited by 421
  3. 2023WinCLIP: Zero-/Few-Shot Anomaly Classification and Segmentationcited by 315
  4. 2022Perception Prioritized Training of Diffusion Modelscited by 303
  5. 2023Effective Data Augmentation With Diffusion Modelscited by 283
  6. 2023FreeU: Free Lunch in Diffusion U-Netcited by 196
  7. 2021The MVTec 3D-AD Dataset for Unsupervised 3D Anomaly Detection and Localizationcited by 178
  8. 2022DeSTSeg: Segmentation Guided Denoising Student-Teacher for Anomaly Detectioncited by 154
  9. 2023Multimodal Industrial Anomaly Detection via Hybrid Fusioncited by 139
  10. 2022Registration based Few-Shot Anomaly Detectioncited by 133
  11. 2022Diffusion Probabilistic Model Made Slimcited by 121
  12. 2021Self-Supervised Predictive Convolutional Attentive Block for Anomaly Detectioncited by 119
16 more
  1. 2022Asymmetric Student-Teacher Networks for Industrial Anomaly Detectioncited by 117
  2. 2022Back to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detectioncited by 93
  3. 2025Marigold: Affordable Adaptation of Diffusion-Based Image Generators for Image Analysiscited by 78
  4. 2023Towards Scalable 3D Anomaly Detection and Localization: A Benchmark via 3D Anomaly Synthesis and A Self-Supervised Learning Networkcited by 67
  5. 2024R3D-AD: Reconstruction via Diffusion for 3D Anomaly Detectioncited by 45
  6. 2023EasyNet: An Easy Network for 3D Industrial Anomaly Detectioncited by 43
  7. 2023Multimodal Industrial Anomaly Detection by Crossmodal Feature Mappingcited by 42
  8. 2023Cheating Depth: Enhancing 3D Surface Anomaly Detection via Depth Simulationcited by 33
  9. 2024Look Inside for More: Internal Spatial Modality Perception for 3D Anomaly Detectioncited by 30
  10. 2022Let Images Give You More:Point Cloud Cross-Modal Training for Shape Analysiscited by 24
  11. 2024Self-supervised Feature Adaptation for 3D Industrial Anomaly Detectioncited by 16
  12. 2024Revisiting Multimodal Fusion for 3D Anomaly Detection from an Architectural Perspectivecited by 12
  13. 2023REB: Reducing Biases in Representation for Industrial Anomaly Detectioncited by 10
  14. 2024Incomplete Multimodal Industrial Anomaly Detection via Cross-Modal Distillationcited by 6
  15. 2025Commonality in Few: Few-Shot Multimodal Anomaly Detection via Hypergraph-Enhanced Memorycited by 6
  16. 20252D_3D Feature Fusion via Cross-Modal Latent Synthesis and Attention Guided Restoration for Industrial Anomaly Detectioncited by 4

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Cross-Modal Mapping and Dual-Branch Reconstruction for 2D-3D Multimodal Industrial Anomaly Detection
  2. 2026Rethinking Auxiliary Modalities in Multimodal Zero-shot Anomaly Detection: From Semantic Fusion to Conditional Modulation
  3. 2025InvAD: Inversion-based Reconstruction-Free Anomaly Detection with Diffusion Models
  4. 2025Correcting Deviations from Normality: A Reformulated Diffusion Model for Multi-Class Unsupervised Anomaly Detection
  5. 2024Enhancing Multi-Class Anomaly Detection via Diffusion Refinement with Dual Conditioning
  6. 2024Dual-Interrelated Diffusion Model for Few-Shot Anomaly Image Generation
  7. 2023DiAD: A Diffusion-based Framework for Multi-class Anomaly Detection
  8. 2023TransFusion -- A Transparency-Based Diffusion Model for Anomaly Detection