Local-global normality learning and discrepancy normalizing flow for unsupervised image anomaly detection

Haiming Yao, Wei Luo, Weihang Zhang, Xiaotian Zhang, Zhenfeng Qiang, Donghao Luo

Abstract: Highlights•This paper introduces a novel two-stream approach for unsupervised image anomaly detection.•Dual-branch Transformer is proposed for local–global reconstruction.•Discrepancy normalizing flow is proposed for defect discrimination.•The proposed method is validated on benchmarks and achieves superior performance.