A5031800506- Face recognition and analysis
- Biometric Identification and Security
- Generative Adversarial Networks and Image Synthesis
- Video Surveillance and Tracking Methods
- Advanced Image Processing Techniques
- Advanced Vision and Imaging
- Experimental Learning in Engineering
- Advanced Control Systems Design
- 3D Shape Modeling and Analysis
- Medical Imaging and Analysis
- Mechatronics Education and Applications
- Gait Recognition and Analysis
Tencent (China)
2024
Beihang University
2018-2021
Shandong Normal University
2021
The generalizable face anti-spoofing (FAS) has attracted much attention recently. Even though many existing methods perform well under intra-domain settings, the model's performance in unseen domain is not satisfying. In this paper, we shift our to frequency seek a solution. Specifically, examine characteristics of different band components FAS images and observe that cross-domain very sensitive low-frequency features. To alleviate sensitivity improve tasks, propose new approach called...
Due to the intrinsic invariance pose and illumination changes, 3D Face Recognition (FR) has a promising potential in real world. FR using high-quality faces, which are of high resolutions with smooth surfaces, have been widely studied. However, research on that low-quality input is limited, although it involves more applications. In this paper, we focus data, targeting an efficient accurate deep learning solution. To achieve this, work two aspects: (1) designing lightweight yet powerful CNN;...
Contemporary face recognition systems use feature templates extracted from images to identify persons. To enhance privacy, template protection techniques are widely employed conceal sensitive identity and appearance information stored in the template. This paper identifies an emerging privacy attack form utilizing diffusion models that could nullify prior protection. The can synthesize high-quality, identity-preserving templates, revealing persons' appearance. Based on studies of model's...
High-level manipulation of facial expressions in images such as expression synthesis is challenging because changes are highly non-linear, and vary depending on the appearance. Identity person should also be well preserved synthesized face. In this paper, we propose a novel U-Net Conditioned Generative Adversarial Network (UC-GAN) for generation. helps retain property input face, including identity information details. We an preserving loss, which further improves performance our model. Both...
Consumer depth sensors have become increasingly common, however, the data are rather coarse and noisy, which is problematic to delicate tasks, such as 3D face modeling recognition. In this paper, we present a novel lightweight Face Refinement Model (3D-FRM), effectively efficiently improve quality of single facial maps. 3D-FRM has an encoder-decoder structure, where encoder applies depth-wise, point-wise convolutions fusion features different receptive fields capture original discriminative...
Contemporary face recognition systems use feature templates extracted from images to identify persons. To enhance privacy, template protection techniques are widely employed conceal sensitive identity and appearance information stored in the template. This paper identifies an emerging privacy attack form utilizing diffusion models that could nullify prior protection, referred as inversion attacks. The can synthesize high-quality, identity-preserving templates, revealing persons' appearance....
Under Emerging Engineering Education circumstances, this paper aims to cultivate the learning interest and strengthen engineering practice ability of first-year students majoring in automation control by making a self-leveling quadrotor. Firstly, attitude calculation based on complementary filtering PID algorithm quadrotor altitude case is intuitively explained. Furthermore, guides utilize Arduino Uno MPU-6050 implement program step step. Finally, essential hardware circuit diagram are given...