Image target detection model attack method, device, terminal equipment and storage medium

Abstract

The invention discloses an image target detection model attack method. The image target detection model attack method includes the following steps: acquiring a sample image and disturbed image setting information input by a user, wherein the disturbed image setting information includes image size and shape; Image and disturbed image set information to generate disturbed image; generate adversarial samples based on sample image and disturbed image; input adversarial samples into the image target detection model for identification to attack the image target detection model, because the size and shape of the disturbed image are the user’s Self-set, user settings are random and uncertain, therefore, the naked eye is not easy to detect the difference between the adversarial example and the sample image, the disturbance is highly concealed, and the attack is powerful. The invention also discloses an image target detection model attack device, a terminal device and a storage medium, which improves the concealment of the disturbance, thereby increasing the attack force on the image target detection model.

Description

Technical field[0001]The present invention relates to the field of image processing, and more particularly to an image target detection model attack method, device, terminal device, and storage medium.Background technique[0002]With the development of machine learning, image recognition has been continuously improved, and the recognition accuracy is constantly increasing.[0003]However, after adding confrontation, the image target detection model can be made completely different results. Therefore, in order to improve the robustness of the image target detection model, it is often necessary to enhance the image target detection model by counterproduction to enhance the robustness of the image target detection model. Different confrontation samples have different attack effects, the greater the attack, the more robust the image target detection model after the attack. Therefore, the choice of counterming samples has a crucial effect on the robustness of the image target detection model...

AI Technical Summary

Problems solved by technology

[0005] The main purpose of the present invention is to provide an image target detection model attack method, device, terminal equipment and storage medium, aiming to solve the technical problem of low attack strength of existing adversarial samples

Images