Abstract
The computational processes in the intermediate stages of the ventral pathway responsible for visual object recognition are not well understood. A recent physiological study by Pasupathy and Connor (2001) in intermediate area V4 using contour stimuli, proposes that a population of V4 neurons display object-centered, position-specific curvature tuning. The standard model of object recognition, developed by Riesenhuber and Poggio (1999) to account for recognition properties of IT cells (extending classical suggestions by Hubel, Wiesel and others, and incorporating standard findings and assumptions about the architecture of the ventral pathway), is used here to model the response of the V4 cells. The model is a feedforward processing hierarchy of increasing invariance and specificity, necessary for object recognition. Our results show that a simple network-level mechanism can exhibit selectivity and invariance properties that correspond to the responses of the V4 cells described by Pasupathy and Connor. These results suggest how object-centered, position-specific curvature tuning of V4 cells may arise from combinations of complex V1 cell responses. Also see the abstract by Serre and Poggio (VSS, 2005) on how such tuning may be learned through a biologically plausible learning mechanism. Furthermore, the model makes predictions about the responses of the same V4 cells studied by Pasupathy and Connor to additional classes of stimuli, such as gratings and natural images. These predictions suggest specific experiments to further explore shape representation in V4.
The authors would like to thank A. Pasupathy and C. Connor for many useful comments and discussions.