felekekinfe/Amharic-Character-Recognition-With-Convolutional-Neural-Networks

# 🌟 **Amharic Hand written Letter Recognition System** ## 📖 Overview Welcome to the **Amharic Hand written Letter Recognition System**! This project uses a **Convolutional Neural Network (CNN)** built with **TensorFlow** to classify Amharic characters from images. Amharic, the official language of Ethiopia, uses the unique Ge'ez script. Our system recognizes 22 Amharic letters, each in all 7 of their forms, totaling **154 unique characters**. This system is a powerful tool for: - 📝 **Optical Character Recognition (OCR)** for Amharic texts - 📚 **Language learning tools** - 🖼️ **Digitizing handwritten Amharic documents** With this project, you get a complete pipeline, from data preprocessing to model training and prediction, all tailored for recognizing Amharic characters. --- ## ✨ Features - **Data Preprocessing**: Seamlessly preprocesses Amharic character images for prediction. - **CNN Model**: Custom-built Convolutional Neural Network for classifying Amharic characters using **TensorFlow**. - **Training & Evaluation**: Tools to train and evaluate the model. - **Prediction**: Predict Amharic characters from new images. --- ## 🗂️ Project Structure Here’s how the project is organized: ``` 📁 dataset/ └── train/ ├── ሀ/ # Images of ሀ ├── ሁ/ # Images of ሁ ├── ሂ/ # Images of ሂ ├── ሃ/ # Images of ሃ ├── ሄ/ # Images of ሄ ├── ህ/ # Images of ህ ├── ሆ/ # Images of ሆ ├── ለ/ # Images of ለ ├── ሉ/ # Images of ሉ ├── ሊ/ # Images of ሊ ├── ላ/ # Images of ላ ├── ሌ/ # Images of ሌ ├── ል/ # Images of ል ├── ሎ/ # Images of ሎ ├── ... # (Folders for the remaining letters) ``` - **`dataset/`**: Contains your training data. - **`model_construction/`**: Includes the CNN model architecture. - **`src/`**: Core scripts for data loading, training, prediction, and preprocessing. - **`trash/`**: Miscellaneous or temporary files. - **`.gitignore`**: Specifies files to ignore in version control. - **`h.py`**: Helper functions (if any). - **`word_list.txt …