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

最新推荐文章于 2024-11-09 21:51:15 发布

Best_Me07

最新推荐文章于 2024-11-09 21:51:15 发布

阅读量1k

点赞数 10

分类专栏：图像异常检测文章标签：算法图像处理开发语言计算机视觉目标检测

本文链接：https://blog.csdn.net/qq_52291558/article/details/142740778

版权

图像异常检测专栏收录该内容

16 篇文章 1 订阅

订阅专栏

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.

研究进展：

基于 CNN 和基于 Transformer 的方法： 是有效的。
进一步讲 CNN 和 Transformer 方法的缺陷：
- CNN： 不好处理远程依赖。
- Transformer： 二次计算复杂度。

Mamba-based models, with their superior long-range modeling and linear efficiency, have garnered substantial attention.
提到新模型：基于 Mamba 的模型

优点： 卓越的远程建模 + 线性效率

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 proposed LSS module, integrating parallel cascaded (Hybrid State Space) HSS blocks and multi-kernel convolutions operations, effectively captures both long-range and local information. The HSS block, utilizing (Hybrid Scanning) HS encoders, encodes feature maps into five scanning methods and eight directions, thereby strengthening global connections through the (State Space Model) SSM.
论文工作：

创新点： Mamba 应用多类无监督异常检测
MambaAD：
- 预训练的编码器
- 具有（局部增强状态空间）LSS 模块
  - 并行级联（混合状态空间）HSS 块
    - 做的： HSS 块利用(混合扫描)HS 编码器
    - 结果： 将特征映射编码为5种扫描方法和8个方向
    - 功能： 通过(状态空间模型)SSM加强全局连接
  - 多核卷积操作
    - 功能： 有效地捕获远程和本地信息
- Mamba 解码器

组成多尺度

The use of Hilbert scanning and eight directions significantly improves feature sequence modeling.
Hilbert：

做到： 扫描和使用八个方向
效果： 显着提高了特征序列建模

Comprehensive experiments on six diverse anomaly detection datasets and seven metrics demonstrate state-of-the-art performance, substantiating the method’s effectiveness.
验证方法有效性：