PBIO SEMINAR SERIES

The neuronal representation underlying visual object recognition in primates

Wednesday - January 25, 2006
05-06 SEMINAR SERIES

James DiCarlo
Asst. Professor MIT
Speaker's website

Host: Bharathi Jagadeesh

The long-term goal of our research is an understanding of the neuronal computations that support the brain’s remarkable ability to recognize visual objects. The key computational challenge of object recognition is the extraction of object identity irrespective of visual clutter, object position, size, pose and illumination. Our working hypothesis is that a series of brain processing stages rapidly transform each pixel-based image of the world into a pattern of neuronal activity that emphasizes object identity and discounts object position, size, view, and illumination. Such a neuronal representation would be well-suited to directly support a range of visual recognition tasks, including report of object identity and category. Because several lines of evidence suggest that this neuronal representation is conveyed by the inferotemporal cortex (IT), our ongoing studies are focused on understanding the IT neuronal population representation, how it is produced, and its role in supporting recognition behavior.
We have recently characterized the suitability of the IT population representation for supporting position- and scale-tolerant recognition. Although several mechanistic hypotheses may explain the remarkable tolerance properties of the IT representation, one of the most intriguing is the possibility that visual experience plays a strong role in shaping such tolerance. In this talk, I will present results from our ongoing studies aimed at testing this hypothesis. These studies illuminate the potential role of the IT representation in supporting visual object recognition, and provide new constraints on the mechanisms that might produce the IT representation.