Image conversion method and device, computer equipment and storage medium

Abstract

The invention relates to an image conversion method and device, computer equipment and a storage medium. The method comprises the steps: obtaining a first image containing face information of a to-be-converted object; inputting the first image into a trained image conversion model, wherein the image conversion model is used for extracting facial features of a to-be-converted object in the first image and generating a second image of the to-be-converted object based on the facial features, and the second image and the first image have different image styles; obtaining a target image corresponding to the first image based on a second image output by the image conversion model, wherein the image conversion model comprises a generative adversarial network, the generative adversarial network comprises a generator and a discriminator, and the generator is used for extracting facial features of the to-be-converted object and generating a second image based on the facial features, and the discriminator is used for assisting the generator to gradually transit from generation of a rough second image to generation of a high-quality second image in the training process of the image conversionmodel. By adopting the method, the accuracy of image conversion can be improved.

Description

technical field [0001] The present application relates to the field of computer vision technology, and in particular, to an image conversion method, apparatus, computer equipment and storage medium. Background technique [0002] With the continuous development of computer vision technology, image conversion technology has gradually become a research hotspot in the field of artificial intelligence, and is widely used in various social applications and website platforms. The shooting style is converted to anime style, that is, the face is animated. [0003] The traditional image transformation method based on machine learning usually uses atrous convolution or deformable convolution to enhance the receptive field of the discriminator in the machine learning network to the image, which solves the problem of image transformation and deformation but does not consider the difference between image features. As a result, the image conversion quality is not high and the accuracy is ...

AI Technical Summary

Problems solved by technology

[0003] Image conversion methods based on machine learning in traditional technologies usually use dilated convolution or deformable convolution to enhance the receptive field of the discriminator in the machine learning network, which solves the problem of image conversion deformation but does not consider the relationship between image features. , so that the image conversion quality is not high and the accuracy is low

[0004] Therefore, the image conversion method in the traditional technology has the problem of low image conversion accuracy

Images