SpikeNet: real-time visual processing with one spike per neuron

Abstract SpikeNet is an image-processing system that uses very large-scale networks of asynchronously firing neurons. The latest version allows very efficient object identification in real-time using a video input, and although this specific implementation is designed to run on standard computer hardware, there are a number of clear implications for computational neuroscience. Specifically, SpikeNet demonstrates the plausibility of visual processing based on a single feed-forward pass and very sparse levels of firing. Above all, it is one of the very few models compatible with the severe temporal constraints imposed by experimental data on processing speed in the visual system.