[Probcogsci] interesting talk

[Angela J. Yu]

Some of you may be interested in this talk on Monday at 2 PM:

SUN: Top-down visual saliency using natural statistics

When people try to find particular objects in natural scenes they make  
extensive use of top-down cognitive knowledge.  Although many forms of  
top-down knowledge have been incorporated into saliency map models of  
visual search, the role of object appearance has been infrequently  
investigated. Here we present an appearance based saliency model  
derived in a Bayesian framework. We compare our model's ability to  
predict where people will look in images when they are told to search  
for a particular object with bottom-up saliency algorithms as well as  
another top-down algorithm known as the Contextual Guidance model.   
Although both top-down approaches use very different types of  
information, they achieve similar performance, and both substantially  
exceed purely bottom-up models that do not take the current task into  
account. Our experiments reveal that the simple model of object  
appearance we present can predict human fixations quite well, even  
making the same mistakes as people.


For more information about the talk and the series:
http://www.cs.ucsd.edu/users/elkan/259/


---------------------------------------------------------
Angela J. Yu
Assistant Professor
Department of Cognitive Science
UCSD, Mail Code 0515
9500 Gilman Drive
La Jolla, CA 92093-0515

Email: ajyu at cogsci.ucsd.edu
Phone: 858-822-3317
Fax: 858-534-1128
Website: www.cogsci.ucsd.edu/~ajyu
---------------------------------------------------------