University of Minnesota
University Relations
http://www.umn.edu/urelate
612-624-6868
myU OneStop


Go to unit's home.

Home | Seminars and Symposia | Past seminars/symposia: Tuesday, April 15, 2014

DTC Science and Technology Innovators Lecture Series

Active Inference of Representations: Control's Role in Visual Perception and Vice-versa

by

Stefano Soatto
Computer Science Department
UCLA

Tuesday, April 15, 2014
3:30 p.m. reception
4:00 p.m. lecture

401/402 Walter Library

Stefano Soatto

The state of a system or agent, understood as a function of measured data that is useful towards a control or decision task, should ideally separate sensing and control: Sensing would infer the function of all past data that is sufficient and hand it off to a control or decision module — agnostic of how the state or representation is inferred — to accomplish the task. While this is indeed possible for linear systems in Gaussian noise, complex sources of uncertainty make the separation imperfect if not impossible. Specifically, when uncertainty is due to sensing mechanisms that involve occlusion and scaling — such as visual sensing, whether in the visible or other spectra — control is actually necessary to infer a state that is sufficient to accomplish even elementary decision tasks. In addition, there may be uncertainty on the task itself. In this talk, we will explore ways of formalizing the properties that an ideal representation should have to support a variety of decision, control and interaction tasks with physical space, where sensing is provided by visual as well as other modalities. We will then see how some drastic simplifications yields to methods that are currently in use today, and point to ways to improve them. We will show applications in visual recognition (finding a known object in an unknown environment) as well as reconstruction (building a model of the environment to support navigation tasks) exploiting visual and inertial sensors.

 

Stefano Soatto is the founder and director of the UCLA Vision Lab (vision.ucla.edu). He received his Ph.D. in Control and Dynamical Systems from the California Institute of Technology in 1996; he joined UCLA in 2000 after being Assistant and then Associate Professor of Electrical and Biomedical Engineering at Washington University, Research Associate in Applied Sciences at Harvard University, and Assistant Professor in Mathematics and Computer Science at the University of Udine, Italy. He received his D.Ing. degree (highest honors) from the University of Padova—Italy in 1992. Dr. Soatto is the recipient of the David Marr Prize (with Y. Ma, J. Kosecka and S. Sastry) for work on Euclidean reconstruction and reprojection up to subgroups. He also received the Siemens Prize with the Outstanding Paper Award from the IEEE Computer Society for his work on optimal structure from motion (with R. Brockett). He received the National Science Foundation Career Award and the Okawa Foundation Grant. He is a Member of the Editorial Board of the International Journal of Computer Vision (IJCV), the International Journal of Mathematical Imaging and Vision (JMIV) and Foundations and Trends in Computer Graphics and Vision.