资 源 简 介
基于神经网络的文字识别系统
本目录主要包括:文档和两个源代
其中源代码是识别程序的,另一个是矩阵类库的。已经保存了训练好的网络权值,所以第一次使用识别程序时,可以直接运行Release目录下的可执行文件,然后对图片目录中的测试图片进行读入、识别。当然也可以自行用训练样本训练网络,不过要注意训练样本的选择,否则可能识别率很低。训练样本选择的原则是,尽可能的有代表性,在训练时间不至于太长的情况下训练样本数目尽量多。-Based on neural network character recognition system catalog include: the two source documents and source code generation which is to identify the procedure, and the other is the matrix class library. Has been preserved trained network weights, so the identification procedures used for the first time, you can directly run the Release directory of the executable file, and then on the picture of the test directory to read into the picture, identification. Of course, we can use training samples to train the network, but attention should be paid to training samples, otherwise it may be very low recognition rate. Training sample selection principle is that, as far as possible have representation in the training time not too long a number of cases as many as possible training samples