Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/shervinnd/persian_alphabet_detection

🧠 Persian_Alphabet_Detection is a deep learning project for recognizing Persian alphabet characters using Dense & CNN models in Keras/TensorFlow. It includes dataset preprocessing, model training, accuracy and ROC curve visualization, and real-world testing for 43 distinct character classes.
https://github.com/shervinnd/persian_alphabet_detection

computer-vision convolutional-neural-networks data-visualization deep-learning image-classification keras machine-learning neural-networks persian-nlp python

Last synced: 11 months ago
JSON representation

🧠 Persian_Alphabet_Detection is a deep learning project for recognizing Persian alphabet characters using Dense & CNN models in Keras/TensorFlow. It includes dataset preprocessing, model training, accuracy and ROC curve visualization, and real-world testing for 43 distinct character classes.

Awesome Lists containing this project

README 

          
# **Persian_Alphabet_Detection**📚✨



Welcome to the **Alphabet Recognition** project! This repository

implements a deep learning-based solution for recognizing Persian

alphabet characters using both **Fully Connected Neural Networks** and

**Convolutional Neural Networks (CNNs)**. Built with Python and Keras,

this project showcases the power of machine learning in image

classification tasks. 🚀



------------------------------------------------------------------------



## 📖 Overview



This project focuses on classifying Persian alphabet characters from

images using two distinct neural network architectures:



-   **Model 1**: A fully connected neural network with multiple dense

    layers.

-   **Model 2**: A convolutional neural network (CNN) with

    convolutional, batch normalization, max-pooling, and dropout layers.



The models are trained and evaluated on custom datasets, with additional

testing on real-world images. The project includes data preprocessing,

model training, performance evaluation, and visualization of results

like ROC curves and sample predictions. 📊



------------------------------------------------------------------------



## 🛠️ Features



-   **Data Preprocessing**: Custom `DataLoader` class for loading,

    resizing, normalizing, and augmenting images (zoom, invert, etc.).

    🖼️

-   **Model Architectures**:

    -   **Model 1**: Sequential dense layers with ReLU activation and

        softmax output for 43 classes.

    -   **Model 2**: CNN with Conv2D, BatchNormalization, MaxPooling,

        and Dropout for robust feature extraction.

-   **Training & Evaluation**:

    -   Trained on three datasets (`DS-01`, `DS-02`, `DS-03`) with

        validation splits.

    -   Performance metrics: Accuracy, Loss, ROC curves, and AUC for

        each class.

-   **Visualization**:

    -   Sample images from training and test sets.

    -   Training/validation accuracy and loss plots.

    -   ROC curves for each class in a 7x7 grid.

    -   Real-world image predictions with side-by-side model

        comparisons.

-   **Real-World Testing**: Preprocesses and predicts on real-world

    images using both models. 🌍



------------------------------------------------------------------------



## 📂 Project Structure



``` plaintext

Alphabet-Recognition/

├── Datasets/

│   ├── DS-01/                # Dataset 1

│   ├── DS-02/                # Dataset 2

│   ├── DS-03/                # Dataset 3

│   └── Real Data/            # Real-world test images

├── Models/

│   └── Neural Network/       # Saved model files

├── Datasets/

│   └── DataLoader/           # DataLoader class

├── alphabet_recognition.py    # Main script

└── README.md                 # Project documentation

```



------------------------------------------------------------------------



## 🚀 Getting Started



### Prerequisites



-   Python 3.8+ 🐍



-   Required libraries:



    ``` bash

    pip install keras matplotlib numpy pandas scikit-learn opencv-python

    ```



### Installation



1.  Clone the repository:



    ``` bash

    git clone https://github.com/shervinnd/Alphabet-Recognition.git

    cd Alphabet-Recognition

    ```



2.  Install dependencies:



    ``` bash

    pip install -r requirements.txt

    ```



3.  Download or prepare the datasets (`DS-01`, `DS-02`, `DS-03`, and

    `Real Data`) and place them in the `Datasets/` folder.



### Running the Project



1.  Update the dataset paths in `alphabet_recognition.py` to match your

    local setup:



    ``` python

    DATASET1 = "path/to/DS-01"

    DATASET2 = "path/to/DS-02"

    DATASET3 = "path/to/DS-03"

    REAL_DATA = "path/to/Real Data"

    ```



2.  Run the main script:



    ``` bash

    python alphabet_recognition.py

    ```



------------------------------------------------------------------------



## 📈 Model Details



### Model 1: Fully Connected Neural Network



-   **Architecture**:

    -   Input: 64x64 grayscale images

    -   Layers: Flatten → Dense (2048, 1024, 512, 256, 64, 43) with ReLU

        and softmax

-   **Training**: Adam optimizer, sparse categorical crossentropy loss,

    20 epochs

-   **Performance**: Evaluated with accuracy, loss, and ROC curves



### Model 2: Convolutional Neural Network



-   **Architecture**:

    -   Input: 64x64 grayscale images (reshaped to 1x64x64)

    -   Layers: Conv2D → BatchNorm → MaxPooling → Dropout (x2) → Flatten

        → Dense (512, 256, 43)

-   **Training**: Adam optimizer, sparse categorical crossentropy loss,

    20 epochs

-   **Performance**: Higher accuracy due to convolutional feature

    extraction



------------------------------------------------------------------------



## 📊 Results



-   **Training/Validation Accuracy & Loss**:

    -   Visualized for both models using Matplotlib.

-   **Test Accuracy**:

    -   Model 1: \~\[Insert test accuracy from script\]

    -   Model 2: \~\[Insert test accuracy from script\]

-   **ROC Curves**:

    -   Plotted for all 43 classes, showing AUC for each.

-   **Real-World Predictions**:

    -   Both models predict on real-world images, displayed side-by-side

        for comparison.



------------------------------------------------------------------------



## 🖼️ Visualizations



### Sample Images



The script visualizes one image per class from the training set and 10

test images with true and predicted labels.



### Training Plots



Training and validation accuracy/loss are plotted to analyze model

performance over epochs.



### ROC Curves



A 7x7 grid of ROC curves shows the performance of each class, with AUC

values for detailed insights.



------------------------------------------------------------------------



## 🔧 Usage



1.  **Training**: Run `alphabet_recognition.py` to train both models and

    generate visualizations.

2.  **Testing**: The script evaluates models on test data and real-world

    images.

3.  **Customization**:

    -   Adjust `EPOCHS`, `IMAGE_SIZE`, or `SHRINK` in the script for

        experimentation.

    -   Modify the `DataLoader` parameters (e.g., `zoom`, `contrast`)

        for different preprocessing.



------------------------------------------------------------------------



## 🤝 Contributing



Contributions are welcome! 🙌 To contribute:



1.  Fork the repository.

2.  Create a new branch (`git checkout -b feature/your-feature`).

3.  Commit your changes (`git commit -m "Add your feature"`).

4.  Push to the branch (`git push origin feature/your-feature`).

5.  Open a Pull Request.



------------------------------------------------------------------------



## 📜 License



This project is licensed under the MIT License. See the LICENSE file for

details.



------------------------------------------------------------------------



## 🙏 Acknowledgments



-   Inspired by Persian alphabet recognition challenges.

-   Thanks to the open-source community for libraries like Keras,

    Matplotlib, and OpenCV.

-   Dataset credits: \[Insert dataset source or credit if applicable\].



------------------------------------------------------------------------



## 📬 Contact



For questions or feedback, reach out via GitHub Issues or connect with

the project maintainer at \[shervindanesh8282@gmail.com\].



Happy coding! 🎉