Abstract

Knowledge distillation (KD) has been widely studied in unsupervisedIndustrial Image Anomaly Detection (AD), but its application to unsupervisedmultimodal AD remains underexplored. Existing KD-based methods for multimodalAD that use fused multimodal features to obtain teacher representations facechallenges. Anomalies in one modality may not be effectively captured in thefused teacher features, leading to detection failures. Besides, these methodsdo not fully leverage the rich intra- and inter-modality information. In thispaper, we propose Crossmodal Reverse Distillation (CRD) based on Multi-branchdesign to realize Multimodal Industrial AD. By assigning independent branchesto each modality, our method enables finer detection of anomalies within eachmodality. Furthermore, we enhance the interaction between modalities during thedistillation process by designing Crossmodal Filter and Amplifier. With theidea of crossmodal mapping, the student network is allowed to better learnnormal features while anomalies in all modalities are ensured to be effectivelydetected. Experimental verifications on the MVTec 3D-AD dataset demonstratethat our method achieves state-of-the-art performance in multimodal anomalydetection and localization.