Training method of image conversion model, image conversion method, device and equipment

By calculating the similarity between the target object mask and the original object mask, and introducing various loss functions and discriminant networks, the image conversion model is updated. This solves the problem that the generated permutation image is not realistic enough when the target image and the original image are significantly different, thus improving the realism of the generated image.

CN116152390BActive Publication Date: 2026-02-06MASHANG CONSUMER FINANCE CO LTD
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202211167836.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-09-23
Publication Date
2026-02-06
Estimated Expiration
2042-09-23

AI Technical Summary

Technical Problem

Existing image transformation methods produce less realistic replacement images when the target image differs significantly from the original image.

Method used

By calculating the similarity between the target object mask and the original object mask, the image conversion model is updated. Multiple loss functions and discriminant networks are introduced to control the training direction of the model and improve the realism of the generated image.

Benefits of technology

The generated permutation images become increasingly realistic, thus improving the realism of the permutation images.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116152390B_ABST
    Figure CN116152390B_ABST
Patent Text Reader

Abstract

The embodiments of the present specification provide a training method of an image conversion model, an image conversion method, an apparatus and a device. The method comprises: inputting an original image with an original object and a target image with a target object into an image conversion model to obtain a replacement image; wherein the replacement image is an image obtained by replacing the original object in the original image with the target object; extracting a target object mask representing the target object from the replacement image, and extracting an original object mask representing the original object from the original image; calculating the similarity between the target object mask and the original object mask; and updating the image conversion model based on the similarity. By calculating the similarity between the mask representing the target object in the replacement image and the mask representing the original object in the original image, and updating the image conversion model based on the similarity, the degree of realism of the replacement image is improved to a certain extent.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Face changing method and device, electronic equipment and storage medium

    CN110889381A