Simple MADALINE network for identification of letters and digits based on ttf file.

This is a simple implementation of a MADALINE network for identification of letters and digits based on a ttf file. The network is trained using the simple single-layer network. The network is implemented in Python using the numpy library for matrix operations.

Usage

1. Create condaenvironment using environment.yml file: conda env create -f environment.yml

2. Activate the environment: conda activate madaline

3. Download the ttf file and put it in data/ directory. Exemplary website. Sample Times Roman font is already included in the repository.

4. Generate the training and test data using ttf file. Example:

   python font_generator.py --width 32 --height 32 --x_position 16 --y_position 16 --font_file data/times-ro.ttf --noise_level 0 --output_dir data/train --overwrite --all_letters```
   
   python font_generator.py --width 32 --height 32 --x_position 16 --y_position 16 --font_file data/times-ro.ttf --noise_level 30 --output_dir data/test_noise_30 --overwrite --all_digits

   You can also use --all_letters or use multiletter string like ABCDEFabcdef to generate only specified letters or digits

5. Train and test the network:

   python madaline_ocr.py --train_path data/train/ --test_path data/test_90

6. Enjoy the results!

Sample results:

python font_generator.py --width 128 --height 128 --x_position 0 --y_position 0 --font_size 32 --font_file data/times-ro.ttf --noise_level 50 --output_dir data/test --overwrite --letter atyukjxZCVF1678

python3 font_generator.py --width 128 --height 128 --x_position 0 --y_position 0 --font_size 32 --font_file data/times-ro.ttf --noise_level 0 --output_dir data/train --overwrite -
-letter atyukjxZCVF1678

python3 madaline_ocr.py --train_path data/train --test_path data/test --plot_results

The sample letters (noisy and clean) are shown below (for the same font and noise level as in the example above):

vs vs

vs vs

TODO

• Add all tests