/Roman-numbers-recognizer

Primary LanguageJupyter NotebookMIT LicenseMIT

Roman-number-recognizer

Description

For handwritten roman numbers recognition task our team creat dataset of roman numbers. It is availabel at link: https://drive.google.com/open?id=1TZF0qVN1PojR5o_nPR_cl7T9sxH37iMY

Saved TensorFlow model is also available at google drive: https://drive.google.com/open?id=1ug2PVwrggSQ7ro2mjGhfXVaZ--j7kLBP

But you don't need manualy download this data. load_and_split_data notebooke will do it for you.

File utils.py contains function for loading data to memory, augmantation and showing images.

Examples of augmanted images are shown in Overview_augmented_images.ipynb.

To run project:

  • install python >= 3.6
  • inslall all packages from requirements.txt in your virtual environment For example, use pip install -r requirements.txt
  • run load_and_split_data.ipynb to load dataset and saved tensorflow model
  • run model1.2_tensorflow.ipynb to see how model works