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.py
provides aVirtualRetina
class and corresponding computation nodes- the
VirtualRetina
class computes the entire cascade of nodes - each node can retain its state, such that input can be chunked
- the
retina_base.py
provides general functions for creating kernelsretina_virtualretina.py
provides default configuration options (as a dictionary) and methods for writing xml configuration files for the original programvison.py
aims at reimplementing the nodes in a more general framework, such that they can be used for any vision computation.