Yaojie Liu

Yaojie Liu

PhD Student @ Michigan State University

About Me

Hi! I'm a second-year Ph.D student at the Department of Computer Science and Engineering, Michigan State University. I was a member of Computational Biology and Cognitive Science Lab (CBCSL), and I'm a current member of Computer Vision Lab (CVLAB), advised by Dr. Xiaoming Liu.

I am broadly interested in computer vision, machine learning and problems related to high-level vision understanding. My current research area is face representation & analysis, including face anti-spoofing, 2D/3D large pose face alignment, and 3D face reconstruction.

Latest Projects

Face Anti-Spoofing

4-year project from IARPA, "Presentation Attack Detection: Solutions for Fingerprints, Face and Iris Systems"

project name

Biometrics utilize physiological, such as fingerprint, face, and iris, or behavioral characteristics, such as typing rhythm and gait, to uniquely identify or authenticate an individual. As biometric systems are widely used in real-world applications including mobile phone authentication and access control, biometric spoof, or Presentation Attack (PA) are becoming a larger threat, where a spoofed biometric sample is presented to the biometric system and attempted to be authenticated. Since face is the most accessible biometric modality, there have been many different types of PAs for faces including print attack, replay attack, 3D masks, etc. As a result, conventional face recognition systems can be very vulnerable to such PAs.

In order to develop a face recognition system that is invulnerable to various types of PAs, in this project, we aim to design a robust face anti-spoofing (or PA de- tection) system to classify a face sample as live or spoof before recognizing its identity.

Real-time demo

project name

DeFA - Dense Face Alignment

DeFA model is designed to provide the dense 3D shape for 2D human faces with large pose and expression variations. We train a CNN using labeled facial landmarks, face contours and SIFT matching as joint supervision. Experiments show our DeFA model does not only provide high-quality dense 3D shape fitting, but also outperforms the state-of-the-art facial landmark detection methods on the challenging datasets. It runs at real time.

Demo on video streams     Related info

project name

Global and Local Fitting of Facial Landmarks and Action Units (AUs)

Facial Landmark Points (FLPs) and facial Action Units show a highly similar distribution among different samples. Hence we propose GL-CNN which provides fast and accurate fine-grained detection and recognition by simultaneous optimization of local and global fits for these two tasks.

Find out more

Publications

Face Anti-Spoofing Using Patch and Depth-Based CNNs

Yaojie Liu, Yousef Atoum, Amin Jourabloo, Arun Ross, Anil Jain, and Xiaoming Liu

International Joint Conference on Biometrics (IJCB), Denver, Colorado, 2017

Fast and Precise Face Alignment and 3D Shape Reconstruction from a Single 2D Image

Ruiqi Zhao, Yan Wang, C. Fabian Benitez-Quiroz, Yaojie Liu, Aleix M. Martinez

European Conference on Computer Vision (ECCV) Workshops, Amsterdam, 2016

Inpainting: Survey and Experiments

Chang Shu, Yaojie Liu

International Journal of Signal Processing, Image Processing and Pattern Recognition, 2016

A Comparison of Image Inpainting Techniques

Yaojie Liu, Chang Shu

Sixth International Conference on Graphic and Image Processing (ICGIP), Beijing, 2014

Dynamic Gesture Localization and Recognition Algorithm Based on Skeletonization

Yaojie Liu, Chang Shu, and Zhizhong Fu

Electronic Science and Technology, 2014