Introduction - If you have any usage issues, please Google them yourself
Help
The first step: Training Network. Training in the use of training samples. (This process may or may not be trained because the training I have saved up a good network parameters, and readers can use to identify)
Step two: identification. First, open the image (256 colors) again, normalized to, click on the "one-time deal" Finally, click "R" or use the menu to find the corresponding item in the Department. Recognition results showed on the screen, as well as output to a file in result.txt.
Recognition rate of the system under normal circumstances was 90 .
In addition, you can open a separate picture of the image pre-processing step by step, but it must be noted that the implementation of each step can only work once, but according to the order of implementation.
Concrete steps as: "256-color bitmap to grayscale"- "grayscale binary"- "noise"- "tip-tilt correction"- "split"- "the standardization of size"- "tight rearrangement."
Note that picture to be