A training method and device of an intrusion warning model and a network attached storage device

Abstract

The application relates to the technical field of intrusion alarm, and discloses a training method and device of an intrusion alarm model and a network-attached storage device. The method comprises the following steps: performing a model layer configuration operation on a basic image comparison model according to image disordering requirements, and obtaining a basic intrusion alarm model; inputting a standard human face image and a training human face image into the basic intrusion alarm model for training, and obtaining an image comparison loss result; judging whether the basic intrusion alarm model meets a model convergence condition according to the image comparison loss result; if yes, it is determined that the basic intrusion alarm model is trained to converge, and the converged basic intrusion alarm model is used for human face image comparison and intrusion alarm. It can be seen that the application can improve the training rationality, accuracy and efficiency of the intrusion alarm model, thereby improving the convergence timeliness and reliability of the intrusion alarm model, and further improving the human face comparison accuracy, intrusion alarm accuracy and timeliness based on the converged intrusion alarm model.

Images