handwritten-digit-recognition

🧠 Handwritten Digit Recognition with CNN

This project demonstrates handwritten digit recognition using a Convolutional Neural Network (CNN) trained on the MNIST dataset.

🚀 Project Structure

• digit_recognition_model.py: Trains a CNN on MNIST and saves the model as .h5
• digit_recognition_model.h5: Trained model file
• prediction.py: Loads the saved model and predicts digits from uploaded images (28x28 grayscale)

🛠️ How to Run

1. Training the Model

Run the training script if you want to retrain the model:

python digit_recognition_model.py

This will create digit_recognition_model.h5.

2. Making Predictions

To predict digits from new images:

python prediction.py

You will be prompted to upload an image (must be 28x28 pixels, grayscale). The script will output the predicted digit along with a plot.

Tip: Use Google Colab if you’re facing hardware limitations.

🧪 Example

🔤 Input Image

The user provides a 28x28 grayscale image of a handwritten digit:

✅ Predicted Output

The model predicts the digit with high accuracy:

🧰 Requirements

• TensorFlow
• NumPy
• matplotlib
• PIL (Pillow)

Install with:

pip install tensorflow numpy matplotlib pillow

📦 Dataset

This project uses the MNIST dataset, which is built into TensorFlow.

📜 License

This project is open-source and free to use under the MIT License.

This site is open source. Improve this page.