Discriminating Animate from Inanimate Visual Stimuli

摘要

From as early as 6 months of age human children distinguish between motion patterns generated by animate objects from patterns generated by moving inanimate objects even when the only stimulus that the child observes is a single point of light moving against a blank background. The mechanisms by which the animate/inanimate distinction are made are unknown, but have been shown to rely only upon the spatial and temporal properties of the movement. In this paper, I present both a multi-agent architecture that performs this classification as well as detailed comparisons of the individual agent contributions against human baselines.