zacharyb02/OCR-tifinagh-script

CNN-based Optical Character Recognition system specifically designed for the Tifinagh script, used for Amazigh languages. This model capable of extracting text from images, scanned documents, then converting it into a usable digital format. # OCR Tifinagh Script **A CNN-based Optical Character Recognition system for Tifinagh script** ## 📖 About This project implements a deep learning-based Optical Character Recognition (OCR) system specifically designed for **Tifinagh script**, the writing system used for Amazigh (Berber) languages. The model can extract text from images, scanned documents, and handwritten content, converting Tifinagh characters into a usable digital format. Tifinagh is used by millions of speakers across North Africa, yet computational support for this script remains limited. This project aims to bridge that gap by providing an accessible, accurate OCR solution for Tifinagh text recognition. ## ✨ Features - **Character Recognition**: Accurately recognizes Tifinagh characters from images - **CNN Architecture**: Leverages deep convolutional neural networks for robust feature extraction - **Multiple Input Formats**: Supports various image formats (PNG, JPG, JPEG) - **Preprocessing Pipeline**: Includes image enhancement and normalization techniques - **Web Interface**: Interactive UI for easy text extraction (HTML/CSS) - **Batch Processing**: Process multiple images efficiently - **High Accuracy**: Achieves strong performance on both printed and handwritten text ## 🚀 Quick Start ### Prerequisites ```bash Python >= 3.8 TensorFlow/Keras NumPy OpenCV Pillow ``` ### Installation 1. **Clone the repository** ```bash git clone github.com cd OCR-tifinagh-script ``` 2. **Install dependencies** ```bash pip install -r requirements.txt ``` 3. **Download the pre-trained model** (if available) ```bash # Place your trained model in the models/ directory ``` ## 💻 Usage ### Web Interface ```bash # Launch the web interface python src/app.py ``` Navigate to `localhost` in your browser to use the interactive OCR interface. ### Jupyter Notebook Explore the training and inference notebooks: ```bash jupyter notebook notebooks/ ``` # …