Description: Digit Recognition

This repository use KNN algorithm provided in OpenCV 3.2 to detect printed numbers. Total line count: 812.

I used 60 different fonts provided by Microsoft Word as input to train the software.
I also rotate the samples [-18, -9, 9, 18] degrees to increase the size of training set.

Use alter in find_main.cpp to detect black digits on white paper or red LED digits on black board.
Threshold parameters can be adjusted in Digit_Finder.h for different light environment.
This software takes only the first digit on the left side if several digits are detected.
It will only confirm the result when a digit is cognized for 8 times continuously.
The software will quit when 10 digits are detected.

TODO

I want to merage the rotating, training and detection all in one main file.
in later OpenCV version, the program may not work well, I will test it when get time.

Name		Name	Last commit message	Last commit date
Latest commit History 17 Commits
Samples		Samples
.gitignore		.gitignore
CMakeLists.txt		CMakeLists.txt
README.md		README.md
digit_finder.cpp		digit_finder.cpp
digit_finder.h		digit_finder.h
find_main.cpp		find_main.cpp
main.cpp		main.cpp
model.yml		model.yml
pipline.png		pipline.png
result.png		result.png
rotate_main.cpp		rotate_main.cpp
train_main.cpp		train_main.cpp
trainer.cpp		trainer.cpp
trainer.hpp		trainer.hpp