Zakariajava/Amazigh-Alphabets-Recognition

# Amazigh Alphabets Recognition This project implements a deep learning model using TensorFlow and Keras to accurately recognize Amazigh alphabet characters from images. The trained model can predict characters from user-supplied images. ## Contents - **`Amazigh-Alphabets-Recognition.ipynb`**: Jupyter Notebook for training the model, visualizing its performance, and saving it. - **`Test_Amazigh-Alphabets-Model.ipynb`**: Python script to load the saved model and make predictions on new images. - **`test_images`**: Directory to store user-provided images for prediction. - **`amazigh_alphabets_model.keras`**: Saved model file containing the trained neural network weights and architecture. ## Model Architecture The model architecture consists of: - **Input Layer**: Conv2D layer with ReLU activation for processing 68x68 RGB images. - **First Convolutional Layer**: 6 filters with 5x5 kernels, followed by MaxPooling. - **Second Convolutional Layer**: 16 filters with 5x5 kernels, followed by MaxPooling. - **Flatten Layer**: Converts the 2D feature maps to a 1D vector. - **Fully Connected Layers**: Dense layers with 120 and 84 units, using ReLU activation. - **Output Layer**: Dense layer with 33 units and softmax activation for multi-class classification (one unit per Amazigh character). Open `Amazigh-Alphabets-Recognition.ipynb` in Jupyter Notebook if you want to add some hidden layer or change the architecture, run all cells to train the model, evaluate its performance, and save it as `amazigh_alphabets_model.keras`. ## Making Predictions To make predictions on new images: 1. Place your images in the `test_images/` folder. Ensure that the images are in PNG format and named appropriately. 2. Run the `Test_Amazigh-Alphabets-Model.ipynb` script. This script will: - Load the saved model from `amazigh_alphabets_model.keras`. - Process the images in the `test_images/` folder. - Display the predictions along with the input images.