Humans have the ability to rapidly and accurately recognize objects in a scene. We perform this task by matching visual inputs with object representations in our memory. These representations are not simply raw images in terms of light and dark pixels, but describe the spatial structure of objects. In many computational models, such representations are implemented as visual schemas, which are active functional units that cooperate and compete to determine which representation best matches the object. This article focuses on how visual schemas can be implemented in neural networks and how they can be used to model human object recognition and scene analysis.

PDF, PS

In M. A. Arbib, editors, The Handbook of Brain Theory and Neural Networks, 1029--1031, Cambridge, MA, 1995. MIT Press.

@inbook{miikkulainen:handbook95, title={Visual Schemas In Object Recognition And Scene Analysis}, author={Risto Miikkulainen and Wee Kheng Leow}, booktitle={The Handbook of Brain Theory and Neural Networks}, editor={M. A. Arbib}, address={Cambridge, MA}, publisher={MIT Press}, pages={1029--1031}, url="http://nn.cs.utexas.edu/?miikkulainen:handbook95", year={1995} }