Abstract

In this paper we introduce a biologically inspired CNN architecture that hasa first convolutional layer that mimics the role of the LGN. The first layer ofthe net shows a rotationally symmetric pattern justified by the structure ofthe net itself that turns up to be an approximation of a Laplacian of Gaussian.The latter function is in turn a good approximation of the receptive profilesof the cells in the LGN. The analogy with respect to the visual systemstructure is established, emerging directly from the architecture of the net.

 

Quick Read (beta)