Invited Speakers

Kristen Grauman CONFIRMED
Kristen Grauman
University of Texas at Austin


Ali Borji
University of Central Florida

On the one hand, wearable devices such as GoPro cameras, smart phones, and glasses have recently provided us with a vast amount of video data from the first-person perspective. Analysis of the videos has quickly become an interesting research area in computer vision with applications such as detecting and recognizing daily activities, video summarization, and localizing the field of view of an egocentric viewer. On the other hand, surveillance cameras and un-manned aerial vehicles capture a significant amount of visual information about daily activities taking place in different locations over long periods of time. Surveillance and generally top view vision has a long history in computer vision; from human detection and re-identification to object tracking. These two types of visual data, although both providing complementary sources of information, have been treated independently in the past. In this disquisition, I will present results of our two recent works. In the first one, we consider a specific scenario which is localizing and identifying people recording the egocentric videos in a top view reference camera (Ardeshir and Borji, ECCV 2016). In the second work, we attempt to transfer motion and action information across egocentric and exocentric domains using deep neural networks (Ardeshir and Borji, In review). We believe that our research is an important first step towards bridging egocentric and surveillance domains, and is useful for many future works and applications such as surveillance, law enforcement, assistive systems, and 3D reconstruction.