Event Location: NSH 1305
Bio: Ajay Mishra is currently working as a Research Scientist at Intelligent Automation Inc, Rockville, Maryland. Prior to this, he was a post-doc/Visiting Researcher working with Prof. Yiannis Aloimonos at the Institute for Advanced Computer Studies, University of Maryland, College Park since 2007. He obtained his PhD (2011) and B.Tech (2003) degrees, both in ECE, from National University of Singapore and Indian Institute of Technology, Kanpur respectively. Before enrolling in the PhD program in 2005, he worked for two years in STMicroelectronics Pvt Ltd. In 2008, he also won the first prize in the Semantic Robot Vision Challenge (Software League).

Abstract: Human perception, being active, is inextricably linked to visual fixation. Despite the obvious importance of fixation, it is still to be an integral part of computer vision/robotics algorithms. Motivated partly by the need to incorporate fixation and attention in a computer vision framework, we proposed a new segmentation algorithm. The algorithm takes a fixation point (i.e a point in the image) as its input and outputs the region corresponding to a distinct surface containing that fixation point (simple objects). All available visual information is utilized in the segmentation process. The fixation points on the objects in the scene are selected through an attentional mechanism that utilizes the knowledge of early border-ownership, also known as figure-ground assignment.

With the attention system, our segmentation framework is fully automatic and segments “simple” objects without any prior knowledge of the scales or the total count of the objects. The output of our segmentation process as an input to the high-level vision algorithms, such as object recognition, object manipulation and action analysis should result in the improved performance of the algorithms.

To see the importance of fixation in visual perception, I advise you to take the psychophysical test available at http://www.umiacs.umd.edu/~mishraka/fixationExperiment.php