22th European Conference on Visual Perception, Perception, 28 (suppl.), pp. 128-129,
August 22-25 - 1999 - Trieste (Italy)
 
Rapid object recognition based on asynchronous feed-forward processing
 
Arnaud Delorme, Rufin van Rullen & Simon Thorpe
 

Centre de Recherche Cerveau & Cognition (UMR 5549) Faculté de Médecine de Rangueil, 133 route de Narbonne, 31062 Toulouse, France.

arno@cerco.ups-tlse.fr , rufin@cerco.ups-tlse.fr, thorpe@cerco.ups-tlse.fr

 
    In humans, 150 ms of processing is sufficient to detect the presence of a target in briefly flashed photographs of natural scenes (Thorpe et al., 1996, Nature, 381, 520), and in monkeys, processing time is probably even shorter (Fabre-Thorpe et al, 1998, Neuroreport, 9, 303). We propose a model for object recognition compatible with this sort of rapid processing in which neurones at each level in a hierarchically organised system fire asynchronously. Starting with the retina, the earliest firing cells are those with the strongest inputs, and at subsequent stages, more complex receptive field properties are produced by a desensitisation mechanism which makes neurones sensitive to the order in which their inputs fire. The final processing layers contain large numbers of neurones trained to particular views of the various targets. Competitive inhibitory mechanisms mean that once one unit has fired, activation of other units with receptive fields in the same part of visual space becomes increasingly difficult. Simulations using SpikeNET, a system designed for modelling very large networks of integrate-and-fire neurones, show that such architectures are remarkably efficient, especially when, for each point in the visual field, there are a large number of feature-selective units in competition.