Experimental evaluation of the CAMERA vision model
The subject of this chapter is a series of experimental tests that were designed to evaluate the CAMERA vision model. Specifically, the evaluation was intended to determine the extent to which an artificial vision system based on the model met the design objectives stated at the beginning of section 9.4. To briefly review the objectives, they stated that the vision system should be capable of being implemented for a reasonable cost, that it should be capable of recognizing relatively complex objects under a variety of conditions, and that it should recognize individual objects in under 5 seconds. The tests were also intended to test the ability of the HAVNET neural network, which is an integral part of the model, to learn and organize object representations. The series of tests, which are described in the following sections, involved the recognition of numerical digits printed black-on-white, puzzle pieces which represent states taken from a wooden puzzle of the United States, and three common three-dimensional objects, a cup, a flashlight and a wrench.
KeywordsRecognition Test Recognition Accuracy Object Representation Camera Movement Camera Model
Unable to display preview. Download preview PDF.