MambaAD: Exploring State Space Models for Multi-class Unsupervised Anomaly Detection

arXiv:2404.06564v1 Announce Type: new
Abstract: Recent advancements in anomaly detection have seen the efficacy of CNN- and transformer-based approaches. However, CNNs struggle with long-range dependencies, while transformers are burdened by quadratic computational complexity. Mamba-based models, with their superior long-range modeling and linear efficiency, have garnered substantial attention. This study pioneers the application of Mamba to multi-class unsupervised anomaly detection, presenting MambaAD, which consists of a pre-trained encoder and a Mamba decoder featuring Locality-Enhanced State Space (LSS) modules at multi-scales. The …

abstract anomaly anomaly detection arxiv attention class cnn cnns complexity computational cs.cv dependencies detection efficiency however linear mamba modeling space state state space models struggle study transformer transformers type unsupervised