ニューラルネットワークによる文字認識

概要

論文の詳細を見る
The purpose of this study is to improve performance of the character recognition tool which use a neural network for, and which enable distinction such as letters from real monuments. The monument offered from Kyoto University has a lot of wounds, and the recognition of the monument's letter is difficult The letter is input into a 50x50 trout in a dot and distinguishes a letter from input by numerousness of the heap of the dot of the mother signal. To perform this effectively, We introduced a function to revise a position and the size of the input signal Furthermore, We introduced the dot input from a picture file to distinguish the letter of the real monument.