| ECVision indexed and annotated bibliography of cognitive computer vision publications This bibliography was created by Hilary Buxton and Benoit Gaillard, University of Sussex, as part of ECVision Specific Action 8-1 The complete text version of this BibTeX file is available here: ECVision_bibliography.bib
G. A. {Takatsuka, M., Caelli, T. M., West} and S. Venkatesh
ABSTRACT
This paper describes a feasibility study of “multi-ageilt oriellted” techiliques 011 a 2-D alld 3-D object recogilitioll system. The mail aim of the project is to develop all illspectioll supportillg tool that ullderstallds objects ill both 2-D alld 3-D ill a ullified system. 2-D alld 3-D worlds are mapped to each other via agent-like entities each of which holds a conceptualized represenitationi, allowing for a robust illferellce ability. Agents, each of which has symbolic represenitationi of a part of a object, are hierarchically organized to represent a complete represenitationi of a object. Ill this paper, object recognitioni is carried out with two matchillg methods: (1) the matchillg betweell a object model (agent’s kniowledge) alld observed data, alld (2) a constraint propagationi-like method to achieve overall conisistanicy among agents. The first is carried out with a symbolic Hopfield-type nieural nietwork alld the seconid via a hierarchical Winmer-Takes-All algorithm. |
|
| Site generated on Friday, 06 January 2006
|