A neural network model of kinetic depth

Abstract

AbstractWe propose a network model that accounts for the kinetic depth in structure from motion phenomena. Using plausible neural mechanisms, the model accounts for (1) fluctuations in perception when viewing a simple kinetic depth stimulus, (2) disambiguation of this stimulus with stereoscopic information, and (3) subsequent bias of the percept of this stimulus following stereoscopic adaptation. The model comprises two levels: a layer of monocular directionally selective motion detectors that provide input to a second layer of disparity- selective and direction-selective binocular mechanisms. The network of facilitatory and inhibitory connections between binocular mechanisms gives rise to fluctuations in network activity that mimic the fluctuations in perception of kinetic depth in the absence of disparity information. The results of a psychophysical experiment are consistent with the nature of the proposed interactions.