"An Examination of Image Recognition with Neural Networks"
Kathy Haines 2005
Abstract
This thesis examines simplified image recognition using artificial neural networks. The images examined were limited to black and white line drawings. Sets of line drawings were analyzed to create a set of input which was fed to a neural network. The neural network was trained to recognize transformed versions of the drawings it learned from the training set. The neural network was designed and trained to learn rotation, scaling, and translation transformations of the original drawings. There appeared to be success learning scaling and translation transformations, with moderate success for rotational transformations. Even though specific restrictions were applied, the goal of the simplified recognition was to create a foundation for more generalized image recognition software.