A simple artificial neural network in the form of a perceptron, which was trained on an open dataset to perform recognition of 26 handwritten letters of the Latin alphabet. The perceptron core part is implemented in two ways: matrix form and graph form.
- GUI: Qt
- Model: C++17
- Install application with:
make install
- Uninstall application with:
make uninstall
- Create distributive:
make dist
- Run tests for model part:
make tests
- Check the code for compliance with Google style:
make style
- Check the code for compliance with cppcheck:
make cppcheck
- Choose between matrix or graph implementation
- Choose number of hidden layers
- Either start learning on a predefined dataset or load weights from file
- Either draw symbol by hand or load *.bmp image from file
- Press classify button
- wmicha (Dmitriy Nikitin)
- hmetallo (Leonid Karataev)
- hsensor (Alexey)