In industrial scenarios, existing deep generative methods face the challenge of adapting to new domains with limited data for anomaly detection. Flow-based generative models are efficient but fail to ...