Unsupervised anomaly detection with a stacked transformer diffusion reconstruction framework

Minjie Du

doi:10.1016/j.eswa.2026.131764

Unsupervised anomaly detection with a stacked transformer diffusion reconstruction framework

2026-02-20

0

PRE

AI

DOI：10.1016/j.eswa.2026.131764

Original

Original request for help

Save

Abstract

En 中文

• Propose a stacked transformer diffusion model for reconstruction-based anomaly detection. • Introduce an Isomorphic Structural Knowledge Guidance module for semantic and geometric alignment. • Employ lightweight Adapter modules to accelerate DiT convergence and enhance efficiency.

Keywords:

anomaly detection

transformer diffusion model

semantic alignment

geometric alignment

lightweight adapter

Unsupervised anomaly detection with a stacked transformer diffusion reconstruction framework

Abstract

Journal

Organization