Desensitization model training method, image desensitization method and device

By using a combination of preset neural networks and desensitization detection networks in image data desensitization, the problem of poor image desensitization in the prior art is solved, and efficient image data desensitization is achieved, personal privacy is protected and information theft is reduced.

CN118862176BActive Publication Date: 2025-03-25ANT GROUP CO LTD
2 Cites 0 Cited by

Patent Information

Application Number
CN202411347168.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-26
Publication Date
2025-03-25
Estimated Expiration
2044-09-26

AI Technical Summary

Technical Problem

In the process of desensitization of image data, it is difficult to effectively improve the image desensitization effect, and it depends on the quality of image text recognition and the quality of sensitive information detection database.

Method used

The preset neural network restores the image after multiple trainings, and combines the desensitization detection network to detect the output image to ensure that the sensitive information in the image is effectively hidden. The method includes acquiring the image sample set, inputting the original image to the preset neural network for image reconstruction, using the desensitization detection network to detect detection errors in the restored image, and training the neural network based on this error until the target desensitization model is obtained.

Benefits of technology

It significantly improves the image desensitization effect, ensures that sensitive information in the image is effectively hidden, protects personal privacy, reduces information theft, and realizes an automated desensitization process, reducing manual intervention.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The embodiments of this specification disclose a desensitization model training method, an image desensitization method and a device. First, an image sample set is obtained, and the image sample set includes at least one original image. Then, each original image is input into a preset neural network for image reconstruction to obtain a corresponding restored image. Furthermore, each restored image is detected using a desensitization detection network to obtain a detection error of a target area in the restored image, and the preset neural network is trained based on the detection error. After the training is completed, a target desensitization model is obtained. Finally, the trained target desensitization model is used to directly perform desensitization processing on the image to be processed.
Need to check novelty before this filing date? Find Prior Art