GANGenerative Adversarial Networks METHOD FOR DETECTING ANOMALY USING GENERATIVE ADVERSARIAL NETWORKS APPARATUS AND SYSTEM THEREOF

摘要

A method for detecting anomaly to improve the accuracy and reliability of a detection result using generative adversarial networks (GAN), an apparatus and a system thereof are provided. An apparatus for detecting anomaly according to some embodiments of the present disclosure may include a memory for storing a GAN-based image transformation model and anomaly detection model, and a processor configured to convert low difficulty learning image into a high difficult learning image through an image transformation model and learning the anomaly detection model using the converted learning image. The anomaly detection model is learned with a high difficulty learning image that is difficult to detect anomalies. So, the detection performance of the apparatus for detecting anomaly can be improved.