Back
Human eye gaze is an important non-verbal cue to estimate the user attention and intention. In this talk, I will present our works over past few years on appearance-based gaze estimation, which takes input frame from a single webcam. I will start with introduction of our work on proposing new datasets such as MPIIFaceGaze, ETH-XGaze, and EVE datasets. And then I will introduce methods such as GazeNet, full face gaze estimation, and gaze redirection. At last, I will briefly introduce applications of our method in real-world settings.
Xucong Zhang (TU Delft)
Assistant Professor
Dr. Xucong Zhang is an assistant professor in the Computer Vision Lab at TU Delft. His research mainly focuses on gaze estimation including datasets, methods, and applications. He is one of the pioneers of appearance-based gaze estimation research leveraging neural networks and taking full face as the input. He established multiple large-scale gaze datasets for benchmarking, where many of the most recent gaze estimation research conducts experiments based on these datasets. His core research interest is human-centered computing as developing techniques for sensing, understanding, and serving the human user. Xucong is now extending his research to broad research areas. Xucong was a postdoc at ETH Zürich working with Prof. Otmar Hilliges. Prior to that, Xucong earned his PhD at Max Planck Institute for Informatics, German, with Prof. Andreas Bulling.
More information