ghurone/mnist-digit-recognizer

A Streamlit app for recognizing handwritten digits

MNIST Digit Recognizer

This Streamlit application allows users to draw and recognize handwritten digits using a trained deep learning model. The app provides an interactive canvas where you can draw digits from 0 to 9 and get real-time predictions with confidence levels.

Features

• Interactive Drawing Canvas: Easily draw digits with a smooth and responsive interface.
• Real-Time Predictions: Get instant predictions for your drawn digits with confidence percentages.
• User-Friendly Interface: Designed with simplicity and ease of use in mind.

How to Use

1. Draw a Digit: Use the canvas to draw any digit from 0 to 9.
2. Predict: Click the 'Predict' button to see the model's prediction and confidence level.
3. Clear and Retry: Easily clear the canvas to draw another digit and make new predictions.

Installation

To run this app locally, follow these steps:

1. Clone this repository:

   git clone https://github.com/ghurone/mnist-digit-recognizer.git

2. Navigate to the project directory:

   cd mnist-digit-recognizer

3. Install the required packages:

   pip install -r requirements.txt

4. Run the Streamlit app:

   streamlit run app.py

Dependencies

• Streamlit
• TensorFlow
• NumPy
• Pillow

Model

The model used in this application is trained on the MNIST dataset, a well-known dataset of handwritten digits. It leverages a Convolutional Neural Network (CNN) to achieve high accuracy in digit recognition.

Acknowledgements

• The MNIST dataset, provided by Yann LeCun, Corinna Cortes, and Chris Burges.
• Streamlit for making it easy to build and share custom web apps for machine learning and data science.

Author

Developed with ❤️ by Erick Ghuron