Computer vision-based anomaly detection method, device and electronic apparatus

A computer-readable storage medium storing computer instructions that cause a computer to perform the following processing:

• dividing a target picture into at least two feature regions according to different region features of the target picture, and forming training sets respectively using the feature regions corresponding to multiple target pictures;
• selecting generative adversarial networks (GAN) as network models to be used, and training GAN network models with the training sets of different feature regions to obtain GAN network models corresponding to different feature regions;
• and
• performing same feature region division on a target picture to be detected, inputting different feature regions of the target picture to be detected into corresponding GAN network models to obtain a generated picture, and performing pixel value-based difference detection on the generated picture and the target picture to be detected,
• wherein dividing a target picture into at least two feature regions according to different region features of the target picture comprises:
• dividing the target picture into a low diversity region and a high diversity region according to diversity of content of the target picture.