Method, system and device for detecting multi-type deep network generated images

Abstract

The invention belongs to the field of deep network generated image, computer vision and image forensics, particularly relates to a method, a system and a device for detecting multiple types of deep network generated images, and aims to solve the problem of low detection rate of the existing deep network generated image detection method for different types of deep network generated images and new types of deep network generated images. The method comprises the following steps: acquiring a to-be-detected image as an input image; preprocessing the input image through a high-pass filter to obtaina preprocessed image; extracting a feature vector of the preprocessed image through a deep residual network; and based on the feature vector, matching the feature vector with the feature vector of each template image in a preset template image library by adopting a template matching method to obtain the category of the to-be-detected image. According to the invention, the detection accuracy of images generated by different types and new types of deep networks is improved.

Description

technical field [0001] The invention belongs to the fields of deep network generated images, computer vision and image forensics, and in particular relates to a method, system and device for detecting multi-type deep network generated images. Background technique [0002] With the advent of the intelligent age, the image can be automatically generated by the computer without the camera. Thanks to the rapid development of Generative Adversarial Network (GAN), the current generation technology based on GAN emerges in an endless stream, and with the continuous optimization of GAN technology, the image quality of the generated image is getting better and better, so that it is almost difficult for human eyes to distinguish between real and fake . Due to the rapid development of deep network image generation technology, software based on image generation technology has emerged as the times require, but some software for editing face images directly affects the news publishing ind...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the above-mentioned problems in the prior art, that is, in order to solve the problem of low detection rate of images generated by existing deep networks for different types and new types of deep networks, the first aspect of the present invention proposes a method for detecting A method for generating images with a multi-type deep network, the method comprising:

Images