由ag国际亚游主办的“信通论坛”本次邀请 Michigan State University的Yaojie Liu博士，与我校师生共同探讨Face Anti-Spoofing。具体安排如下，欢迎感兴趣的师生参加。
一、主 题：Face Anti-Spoofing: Detect, Visualize and Generalize
四、地 点：清水河校区科研楼B区 302会议室
五、主持人：ag国际亚游 舒畅 老师
Face is the most accessible biometric modality which can be used for identity verification, but face recognition system per se is vulnerable to face spoofing or presentation attacks: using spoofing medium (e.g. printed face, digital screen) to present the identification of the target subject and fool the recognition system. Thus, face anti-spoofing is a critical step to prevent face recognition systems from such security breach.
In this talk, we will discuss three aspects of the study of face anti-spoofing: detect, visualize and generalize. Firstly, we will discuss how to build a model that can reliably detect the spoofing faces; Secondly, we will introduce the approach that enables the face anti-spoofing model to provide a visual interpretation and understanding; Thirdly, we will present a method to detect the unknown spoof attacks, namely zero-shot face anti-spoofing.
Yaojie Liu is a fourth-year Ph.D. student and research assistant at Michigan State University in the Computer Vision Lab (CVLab), advised by Prof. Xiaoming Liu. Before joining MSU, he obtained his B.Eng. degree from University of Electronic Science and Technology of China at 2014 and M.Sc. degree from the Ohio State University at 2016. From 2015 to 2016, he was a member of Computational Biology and Cognitive Science Lab (CBCSL), working with Prof. Aleix Martinez. In 2019 summer, he joined Apple AI Research as an intern, working with Dr. Barry Theobald on audio visual modeling. He serves as the reviewer of many international journals and conferences, such as TIP, Neurocomputing, CVPR, AAAI, ICCV, and ECCV. He has one U.S. patent on file.
His research area is computer vision and machine learning, with specific interests in face representation, face alignment, 3D face reconstruction, audio-visual modeling, face anti-spoofing and forensics.