Detect People's Faces and Protect Them by Providing High Privacy Based on Deep Learning

Abstract

Facial privacy is essential in our time due to the violations that occur due to the proliferation of social media and people's primary dependence on it. Facial features can be exploited to identify, track, or other matters without obtaining prior consent. This is increasingly important due to the increasing use of facial recognition technologies. Protecting the face and privacy is a challenging task, as the entire world is very widely connected through social networking sites in an uncensored manner, especially in countries with no electronic governance or oversight. Therefore, there is an urgent need to provide systems or research that focuses on the issue of facial privacy. In this paper, a system for providing privacy for people was proposed using the WIDER FACE data set, considered the most important among the data sets. The system aims to provide privacy for people by determining the destination that must be preserved and provided with privacy through three technical. The approach goes through several steps: The processing process of the image is achieved by enhancement of the images that are input in the training stage and then dividing the data into a test and training set, and the training stage through the YOLOv6 algorithm (looks only once), and privacy operations including encryption, decryption, mask and blurring in the test part of the data, and conducting an external test for personal photo. The final results of the proposed system were as follows: accuracy = 0.98 in training and 0.96 in testing.

Keywords

• deep learning
• facial
• image
• privacy
• YOLOv6