RGB-D Perception: Depth Camera Usages beyond Gesture and Gaming
Xiaofeng Ren, Intel Labs Seattle
RGB-D cameras, i.e. Kinect-style cameras that provide synchronized color and depth, are a rapidly emerging and growing trend of technology. RGB-D cameras lead the way of enabling robust visual perception in everyday scenarios, showing great promises for a wide range of research and applications from robotics to HCI.
In this talk I will mainly discuss two lines of RGB-D research at Intel Labs Seattle in collaboration with the University of Washington. The first line is on 3D mapping and modeling: we develop RGB-D registration algorithms and systems to enable non-technical users to interactively scan indoor spaces into 3D at real time. The second line is on everyday object recognition: we design rich visual features called kernel descriptors and show that (1) they outperform SIFT on major recognition benchmarks; (2) they are flexible and easily adapt, such as to depth images. Both lines of work lead to practical visual perception in real world conditions and show potentials of camera-based rich sensing in consumer applications.
Biography: Xiaofeng Ren is a research scientist at Intel Labs Seattle and also an affiliate assistant professor at the University of Washington. His research interests are broadly in the areas of computer vision and its applications, including image features, grouping and segmentation, object recognition, and video analysis. His current focus is on understanding and solving computer vision problems in everyday life settings. He received his PhD from University of California, Berkeley and his BS from Zhejiang University. Prior to joining Intel in 2008, he was on the research faculty of Toyota Technological Institute at Chicago.
