ManuScript

A Latin manuscript deep learning transcriber

Authors

1. Karen Dianelis Cantero López C-411
2. Luis Alejandro Rodríguez Otero C-411
3. Hector Miguel Rodríguez Sosa C-411
4. Sebastian Suárez Gómez C-411

Introduction

ManuScript is a deep learning project aimed at transcribing Latin manuscripts. The project leverages the power of Optical Character Recognition (OCR) to convert handwritten or printed text in Latin manuscripts into machine-encoded text.

The project uses Python as the primary language and TensorFlow for implementing the deep learning model. The OCR model is trained on a dataset of Italian letters called LAM, allowing it to accurately transcribe the text.

Installation

The project dependencies are listed in the requirements.txt file. To install these dependencies, use the following pip command:

pip install -r requirements.txt

Usage

In the image_preprocessing folder, more precisely the preprocessing_filters.py file, is a script ready to apply different kinds of preprocessing filters to your dataset images. Read the documentation in the file to know how to use it.

Afterward the code relies on all the image dataset chosen for training, validation and testing will remain in the image_preprocessing/Out folder.

All the code necessary to train and evaluate the model is in the model/playground.ipynb notebook.