Towards an Incremental Unified Multimodal Anomaly Detection: Augmenting Multimodal Denoising From an Information Bottleneck Perspective

The quest for incremental unified multimodal anomaly detection seeks to empower a single model with the ability to systematically detect anomalies across all categories and support incremental learning to accommodate emerging objects/categories. Central to this pursuit is resolving the catastrophic forgetting dilemma, which involves acquiring new knowledge while preserving prior learned knowledge. Despite some efforts to address this dilemma, a key oversight persists: ignoring the potential impact of spurious and redundant features on catastrophic forgetting. In this paper, we delve into the negative effect of spurious and redundant features on this dilemma in incremental unified frameworks, and reveal that under similar conditions, the multimodal framework developed by naive aggregation of unimodal architectures is more prone to forgetting. To address this issue, we introduce a novel denoising framework called IB-IUMAD, which exploits the complementary benefits of the Mamba decoder and information bottleneck fusion module: the former dedicated to disentangle inter-object feature coupling, preventing spurious feature interference between objects; the latter serves to filter out redundant features from the fused features, thus explicitly preserving discriminative information. A series of theoretical analyses and experiments on MVTec 3D-AD and Eyecandies datasets demonstrates the effectiveness and competitive performance of IB-IUMAD.

• Publication year

  2026

• Venue

  Unknown venue

• Publication date

  2026-03-03

• Fields of study

  Computer Science

• Identifiers
  arXiv 2603.02629
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Boosting Global-Local Feature Matching via Anomaly Synthesis for Multi-Class Point Cloud Anomaly Detection

  2025cited by this paper
• UniMMAD: Unified Multi-Modal and Multi-Class Anomaly Detection via MoE-Driven Feature Decompression

  2025cited by this paper
• Enhancing Multimodal Learning via Hierarchical Fusion Architecture Search With Inconsistency Mitigation

  2025cited by this paper
• C3D-AD: Toward Continual 3D Anomaly Detection via Kernel Attention with Learnable Advisor

  2025cited by this paper
• UniNet: A Contrastive Learning-guided Unified Framework with Feature Selection for Anomaly Detection

  2025cited by this paper
• Continual-MEGA: A Large-scale Benchmark for Generalizable Continual Anomaly Detection

  2025cited by this paper
• Center-Aware Residual Anomaly Synthesis for Multiclass Industrial Anomaly Detection

  2025cited by this paper
• ReplayCAD: Generative Diffusion Replay for Continual Anomaly Detection

  2025cited by this paper
• MC3D-AD: A Unified Geometry-aware Reconstruction Model for Multi-category 3D Anomaly Detection

  2025cited by this paper
• CostFilter-AD: Enhancing Anomaly Detection through Matching Cost Filtering

  2025cited by this paper
• 3D vision-based anomaly detection in manufacturing: A survey

  2025cited by this paper
• Distribution Prototype Diffusion Learning for Open-set Supervised Anomaly Detection

  2025cited by this paper
• One-for-More: Continual Diffusion Model for Anomaly Detection

  2025cited by this paper
• Background-Weaken Generalization Network for Few-Shot Industrial Metal Defect Segmentation

  2025cited by this paper
• MambaAD: Exploring State Space Models for Multi-class Unsupervised Anomaly Detection

  2024cited by this paper
• SoftPatch: Unsupervised Anomaly Detection with Noisy Data

  2024cited by this paper
• Unsupervised Continual Anomaly Detection with Contrastively-learned Prompt

  2024cited by this paper
• Self-supervised Feature Adaptation for 3D Industrial Anomaly Detection

  2024cited by this paper
• EfficientVMamba: Atrous Selective Scan for Light Weight Visual Mamba

  2024cited by this paper
• Hierarchical Gaussian Mixture Normalizing Flow Modeling for Unified Anomaly Detection

  2024cited by this paper
• A Diffusion-Based Framework for Multi-Class Anomaly Detection

  2024cited by this paper
• Rethinking Reverse Distillation for Multi-Modal Anomaly Detection

  2024cited by this paper
• PromptAD: Learning Prompts with only Normal Samples for Few-Shot Anomaly Detection

  2024cited by this paper
• Learning Feature Inversion for Multi-class Anomaly Detection under General-purpose COCO-AD Benchmark

  2024cited by this paper
• Dinomaly: The Less Is More Philosophy in Multi-Class Unsupervised Anomaly Detection

  2024cited by this paper
• Prior Normality Prompt Transformer for Multiclass Industrial Image Anomaly Detection

  2024cited by this paper
• SeaS: Few-shot Industrial Anomaly Image Generation with Separation and Sharing Fine-tuning

  2024cited by this paper
• Boosting Fine-Grained Visual Anomaly Detection with Coarse-Knowledge-Aware Adversarial Learning

  2024cited by this paper
• Look Inside for More: Internal Spatial Modality Perception for 3D Anomaly Detection

  2024cited by this paper
• Multi-Sensor Object Anomaly Detection: Unifying Appearance, Geometry, and Internal Properties

  2024cited by this paper
• Revisiting Multimodal Fusion for 3D Anomaly Detection from an Architectural Perspective

  2024cited by this paper
• Hierarchical Vector Quantized Transformer for Multi-class Unsupervised Anomaly Detection

  2023cited by this paper
• Shape-Guided Dual-Memory Learning for 3D Anomaly Detection

  2023cited by this paper
• Removing Anomalies as Noises for Industrial Defect Localization

  2023cited by this paper
• An Incremental Unified Framework for Small Defect Inspection

  2023influential reference
• Multimodal Industrial Anomaly Detection via Hybrid Fusion

  2023cited by this paper
• SimpleNet: A Simple Network for Image Anomaly Detection and Localization

  2023cited by this paper
• OmniAL: A Unified CNN Framework for Unsupervised Anomaly Localization

  2023cited by this paper
• Exploring Plain ViT Reconstruction for Multi-class Unsupervised Anomaly Detection

  2023cited by this paper
• Multimodal Industrial Anomaly Detection by Crossmodal Feature Mapping

  2023cited by this paper
• Towards Scalable 3D Anomaly Detection and Localization: A Benchmark via 3D Anomaly Synthesis and A Self-Supervised Learning Network

  2023cited by this paper
• Cheating Depth: Enhancing 3D Surface Anomaly Detection via Depth Simulation

  2023cited by this paper
• Dynamic Multimodal Information Bottleneck for Multimodality Classification

  2023cited by this paper
• Towards Continual Adaptation in Industrial Anomaly Detection

  2022cited by this paper
• The Eyecandies Dataset for Unsupervised Multimodal Anomaly Detection and Localization

  2022cited by this paper
• DSR - A dual subspace re-projection network for surface anomaly detection

  2022cited by this paper
• A Unified Model for Multi-class Anomaly Detection

  2022cited by this paper
• Anomaly Detection via Reverse Distillation from One-Class Embedding

  2022cited by this paper
• DeSTSeg: Segmentation Guided Denoising Student-Teacher for Anomaly Detection

  2022cited by this paper
• Back to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection

  2022cited by this paper
• Asymmetric Student-Teacher Networks for Industrial Anomaly Detection

  2022cited by this paper
• The MVTec 3D-AD Dataset for Unsupervised 3D Anomaly Detection and Localization

  2021cited by this paper
• Natural Synthetic Anomalies for Self-supervised Anomaly Detection and Localization

  2021cited by this paper
• Farewell to Mutual Information: Variational Distillation for Cross-Modal Person Re-Identification

  2021cited by this paper
• CutPaste: Self-Supervised Learning for Anomaly Detection and Localization

  2021cited by this paper
• DRÆM – A discriminatively trained reconstruction embedding for surface anomaly detection

  2021cited by this paper
• Fully Convolutional Cross-Scale-Flows for Image-based Defect Detection

  2021cited by this paper
• Learning Robust Representations via Multi-View Information Bottleneck

  2020cited by this paper
• PaDiM: a Patch Distribution Modeling Framework for Anomaly Detection and Localization

  2020cited by this paper
• Uninformed Students: Student-Teacher Anomaly Detection With Discriminative Latent Embeddings

  2019cited by this paper
• FIND: Few-Shot Anomaly Inspection with Normal-Only Multi-Modal Data

  year unknowncited by this paper

• No citing papers are available for this paper.