Currently I am working in HID group. My research includes detect interesting (unusual) events and human activities in long surveillance video and human body tracking.
My approach to interesting events detection in video employs finding an embedding space of all the events of the video in which the similar events are near to each other and dissimilar events are far away from each other. By doing so, we can automatically find unusual events which are separated from other common events in the space and we can also learn a classifier to detect this event in new videos. This un-supervised approach learns the definition of "unusual" from the data so it has less constrains and assumptions on the input video.
To track human in video, I first sort the video so that frames with similar poses of the human body are put together while those with different body poses are put away from each other. By sorting the video we can come up with some clues of how the body pose changes and how to track it.
In computer graphics, I am particularly interested in computer generated caricatures recently. It is highly artistic to make caricatures. So to let computer draw caricatures is very difficult. I am working on a machine learning and sample-based approach to this difficult problem. This job is done with Microsoft Research Asia in Beijing.
I like playing soccer and we SCS guys have a soccer team. I also like painting and drawings. Taking photograph is also my favorite.
|Research Interest Keywords|
|computer graphics, computer vision, machine learning, visual tracking|
|The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.|
Contact Us | Update Instructions