This python package provides an implementation of the Virtual Retina developed by Adrien Wohrer. It uses theano to simulate spike trains of retinal ganglion cells by directing the input through a number of computation nodes. Each node might do linear or nonlinear computations, eg. convolve the inpute with a spatio-temporal kernel or apply gain control.
retina.pyprovides aVirtualRetinaclass and corresponding computation nodes- the
VirtualRetinaclass computes the entire cascade of nodes - each node can retain its state, such that input can be chunked
- the
retina_base.pyprovides general functions for creating kernelsretina_virtualretina.pyprovides default configuration options (as a dictionary) and methods for writing xml configuration files for the original programvison.pyaims at reimplementing the nodes in a more general framework, such that they can be used for any vision computation.