From even a brief glance of a scene the human visual system achieves a rich interpretation that includes the objects that are present and where they are, the surfaces that compose the space and how they move, and the identity of the individuals that are present and even their cognitive intentions. The visual face processing system provides a unique opportunity to discover the computations and neural basis of this astonishing perceptual ability. It has recently been discovered that the face processing subsystem of macaque visual cortex consists of six discrete cortical patches that are preferentially driven by face stimuli, form a strongly connected subnetwork, and subserve a transformation into neurons selective for identity and robust to visual changes. However, the precise nature of the transformation that occurs within the face patch system and the encoding of face features within each patch is far from understood. In order to understand precisely these processes we seek to develop novel computational models of this transformation and rigorously test these models experimentally. The overall aim of this proposal is to produce testable computational hypotheses of visual face processing and to quantitatively test these hypotheses, along with already existing hypotheses, as models of neural responses in the face patches. The results of this study will provide insight into the principals that govern the transformation of visual information in cortex and will serve as a substrate for the further investigation of normal and abnormal face perception in humans.
Specific cortical mechanisms are involved in visual face perception. We seek to develop computational models and test these models against neural recordings of these specific cortical mechanisms involved in visual face perception. Understanding how faces are processed in cortex may help us understand general information processing in the brain and may lead to treatments and understanding of persons with deficits in visual face perception.
|Cadieu, Charles F; Hong, Ha; Yamins, Daniel L K et al. (2014) Deep neural networks rival the representation of primate IT cortex for core visual object recognition. PLoS Comput Biol 10:e1003963|
|Yamins, Daniel L K; Hong, Ha; Cadieu, Charles F et al. (2014) Performance-optimized hierarchical models predict neural responses in higher visual cortex. Proc Natl Acad Sci U S A 111:8619-24|