header_image

Learning Distances for Retrieval and Detection in Multimedia

Dr. Rahul Sukthankar
Friday, May 25, 2007
11:00AM ~ 12:00PM, Harris Center 125

Abstract


Information retrieval and object detection in multimedia has traditionally focused on identifying appropriate features for the given task. In the past, these were typically manually engineered; currently, they are often generated using machine learning techniques. This talk will discuss how feature selection is closely related to supervised distance metric learning. Specifically, we describe a new method for inducing binary features using a sequence of boosted pairwise classifiers that results in a weighted Hamming distance with good properties. Additionally, these binary features can be viewed as a storage-sensitive method for dimensionality reduction that preserves classification accuracy. We demonstrate applications in several real-world areas including image/video retrieval, music identification and search-assisted diagnosis of medical images.

This research was performed in an open collaborative framework with my research interns, graduate students and faculty colleagues. In particular, I would like to acknowledge Derek Hoiem, Yan Ke, Liu Yang, Rong Jin, Lily Mummert and Bin Zheng.

Short Bio


Rahul Sukthankar is a principal research scientist at Intel Research Pittsburgh and adjunct research professor in the Robotics Institute at Carnegie Mellon. He was previously a senior researcher at HP/Compaq's Cambridge Research Lab and a research scientist at Just Research. Rahul received his Robotics Ph.D. from Carnegie Mellon and his B.S.E. summa cum laude in computer science from Princeton. His current research focuses on computer vision and machine learning, particularly in the areas of object recognition, information retrieval and medical imaging.

FEEDBACK | Webmaster | EECS | FSI | CECS | UCF
University Of Central Florida | Orlando, Florida 32816-2362 Phone: 407-823-2341