Method and device for constructing binary classification anomaly detection

The invention relates to the technical field of computer vision, in particular to a binary classification anomaly detection method and device.Firstly, an open data set is constructed, a binary classification two-way branch model architecture is designed, when the open data set is constructed, a clustering algorithm is used for dividing all normal samples into even number groups which are not intersected with one another, and a binary classification two-way branch model architecture is designed; each group represents one mode of a normal picture, generating a pseudo abnormal sample, and increasing the diversity of abnormal samples; a binary classification double-path branch model architecture adopts a double-path structure and is divided into a basic branch and a correction branch, and the basic branch and the correction branch adopt different data enhancement modes to increase data distribution difference. Compared with the prior art, the method has the advantages that the model can be prevented from overfitt