Dave Andresen
Post-Doctoral Fellow
Neural Bases of View-Invariant Object Recognition
An object can appear very different when seen from various viewpoints. How
such view-invariant object recognition is supported by the visual system
is not well understood. We are using functional magnetic resonance imaging
(fMRI) to examine the regions of visual cortex involved in object
recognition, and examining how viewpoint is represented within these
regions in order to understand the process of view-invariant object
recognition.
Our studies thus far indicate that for some object classes such as
animals, some views elicit more robust activity than other views (see
image). This increased activation appears to be the result of a larger
proportion of neurons within the population being dedicated to
representing these views. Interestingly, more posterior regions such as
LOC are invariant to 60° changes in viewpoint, whereas more anterior
regions along the fusiform gyrus are invariant to more than 120° changes
in viewpoint. These results indicate that objects are represented by
mixtures of view-dependent neural subpopulations within a hierarchy of
increasingly view-invariant object representations. |
|
|
