AEGAN: Generating imperceptible face synthesis via autoencoder‐based generative adversarial network

Abstract

Face recognition (FR) systems based on convolutional neural networks have shown excellent performance in human face inference. However, some malicious users may exploit such powerful systems to identify others' face images disclosed by victims' social network accounts, consequently obtaining private information. To address this emerging issue, synthesizing face protection images with visual and protective effects is essential. However, existing face protection methods encounter three critical problems: poor visual effect, limited protective effect, and trade‐off between visual and protective effects. To address these challenges, we propose a novel face protection approach in this article. Specifically, we design a generative adversarial network (GAN) framework with an autoencoder (AEGAN) as the generator to synthesize the protection images. It is worth noting that we introduce an interpolation upsampling module in the decoder in order to let the synthesized protection images evade recognition by powerful convolution‐based FR systems. Furthermore, we introduce an attention module with a perceptual loss in AEGAN to enhance the visual effects of synthesized images by AEGAN. Extensive experiments have shown that AEGAN not only can maintain the comfortable visual quality of synthesized images but also prevent the recognition of commercial FR systems, including Baidu and iKLYTEK.

Precision in Pursuit: A Multi-Consistency Joint Approach for Infrared Anti-UAV Tracking

“…Deep learning-based visual object tracking can be broadly classified into two major categories: approaches based on convolutional neural networks [30] and those founded on transformer models. Siamese networks initially employ structurally identical [31] but nonweight-sharing feature extraction networks to extract features from template images and search images.…”

Nighttime Driver Behavior Prediction Using Taillight Signal Recognition via CNN-SVM Classifier

“…For data-driven methods to perform well in the real world and challenging conditions, they must be exposed to various environments during the training process [34]. Unfavorable weather, low light conditions, low resolution, dust on lenses, and camera vibration are among these conditions.…”

“…Secondly, utilizing smartphone cameras enhances the model's ability to adapt for implementation in human-driven vehicles, thus increasing its generalization. Thirdly, applying these masks to the dataset is also seen as a form of data augmentation, which helps prevent the model from overfitting during training [34]. Examples of mapping some of these masks at severe levels on the training and testing data are shown in Fig.…”