Event overview
This event is part of the Whitehead Lectures series organized by the Centre for Cognition, Computation and Culture (Department of Psychology), and the Department of Computing
The human capacity for visual object recognition is characterized by a number of properties that are jointly challenging to account for. The visual representation of shape—and thus our ability to recognize objects—is largely invariant with the angle and distance from which an object is viewed. The representation of shape is robust to metric variations in object shape, permitting recognition of novel instances of known object classes. And we perceive an object’s parts and their interrelations independently of one another, permitting us to easily recognize even non-rigid objects. These properties of object recognition are naturally accounted for in terms of an analytic representation of shape that specifies an object’s parts and their (categorical) spatial relations. At the same time, however, object recognition is both fast (feed-forward, at least in the case of over-learned objects in familiar views) and automatic (again, at least for objects in familiar views). The speed and automaticity of object recognition are inconsistent with an account that relies exclusively on (time- and attention-demanding) analytic representations of shape. I will describe a model of object recognition that accounts for these findings in terms of a hybrid analytic/holistic representation of shape. The model makes several novel predictions regarding the relationship between visual attention and patterns of visual priming across different kinds of object images. I will also discuss several experiments testing and verifying these predictions. Together, the properties of visual object recognition and visual priming suggest that the visual system generates analytic representations of attended objects but recognizes unattended objects in familiar views based on rapidly- and automatically-generated holistic representations.
Prof. Hummel received his PhD from the University of Minnesota in 1990 working with Irv Biederman. He took his first job, at UCLA, in 1991 and moved to the University of Illinois in 2005. He has been involved with both empirical and computational work on shape perception, object recognition, analogy, analogical inference, schema induction, cognitive development (namely, computational work on the development and acquisition of relational concepts) and category learning. He is currently working to extend the LISA model of analogical reasoning to provide a process model of explanation (i.e., how people apply their prior knowledge to generate explanations of novel observations or facts).
Dates & times
Date | Time | Add to calendar |
---|---|---|
12 May 2009 | 4:00pm - 6:00pm |
Accessibility
If you are attending an event and need the College to help with any mobility requirements you may have, please contact the event organiser in advance to ensure we can accommodate your needs.