Handwritten Character Recognition using Deep Learning

This project implements a deep learning model for recognizing handwritten characters (both digits and letters) using a Convolutional Neural Network (CNN). The application provides a user-friendly interface for drawing characters and getting real-time predictions.

Features

• Handwritten digit and letter recognition (0-9, A-Z)
• Real-time prediction using a trained CNN model
• Interactive drawing canvas
• Simple and intuitive user interface

Dataset

The model is trained on a combination of two datasets:

1. MNIST Database (Modified National Institute of Standards and Technology database)

   • Contains 60,000 training images and 10,000 testing images
   • Handwritten digits (0-9)
   • 28x28 pixel grayscale images
   • Download: MNIST Dataset

2. EMNIST Dataset (Extended MNIST)

   • Extends MNIST to include handwritten letters (A-Z, a-z)
   • Maintains the same format as MNIST (28x28 pixel grayscale images)
   • Download: NIST EMNIST Dataset

Dataset Setup Instructions

1. Create a directory called sample_data in the project root:

   mkdir sample_data
   

2. Place the following dataset files in the sample_data directory:

   • mnist_train.csv - MNIST training data
   • mnist_test.csv - MNIST test data
   • A_Z Handwritten Data.csv - EMNIST letters data (available from Kaggle or other sources)

3. The directory structure should look like this:

   handwritten-character-recognition-deep-learning/
   ├── sample_data/
   │   ├── mnist_train.csv
   │   ├── mnist_test.csv
   │   └── A_Z Handwritten Data.csv
   ├── model/
   ├── main.py
   ├── training.py
   ├── README.md
   └── requirements.txt
   

Requirements

• Python 3.7+
• TensorFlow 2.x
• Keras
• NumPy
• Pandas
• Pillow
• scikit-learn
• pywin32 (for Windows GUI)
• tkinter (usually comes with Python)

Installation

1. Clone the repository:

   git clone [your-repository-url]

2. Install the required packages:

   pip install -r requirements.txt

Usage

1. Run the training script (optional, if you want to retrain the model):

   python training.py

   Note: Make sure to place the dataset files in the sample_data/ directory:

   • mnist_train.csv
   • mnist_test.csv
   • A_Z Handwritten Data.csv

2. Run the application:

   python main.py

3. Use the application:

   • Draw a character (digit or letter) on the black canvas
   • Click "Predict" to see the model's prediction
   • Click "Clear" to clear the canvas and start over

Model Architecture

The model uses a deep CNN with the following layers:

• Multiple Conv2D layers with BatchNormalization
• MaxPooling2D for downsampling
• Dropout layers for regularization
• Dense layers for classification
• Softmax activation for output (36 classes: 0-9, A-Z)

Performance

The model achieves high accuracy on both training and test sets, with data augmentation techniques applied to improve generalization.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is open source and available under the MIT License.

Acknowledgments

• MNIST Database
• EMNIST Dataset
• Keras and TensorFlow teams for the deep learning framework