live demo: https://digit-recognition-webapp.onrender.com/
It may take up to 1 minute to load the website if the server is off.
This project demonstrates the development of a deep learning-based web application using Flask and TensorFlow for handwritten digit recognition. The application involves two main components:
- Model Training (training.py):
The project trains a Convolutional Neural Network (CNN) on the MNIST dataset, a widely used benchmark for handwritten digit recognition. The key features of the training script include: A convolutional layer with 32 filters and ReLU activation, A max-pooling layer for dimensionality reduction and an output layer with a softmax activation for multi-class classification of digits (0–9).
- Web Application (app.py):
The Flask web application serves as an interface for real-time digit prediction using the trained model. Accepts JSON input containing a 28x28 grayscale image matrix. Reshapes the matrix into a format compatible with the CNN (1x28x28x1). Performs prediction using the trained model and returns the predicted digit as JSON response.
pip install flask
pip install numpy
pip install tensorflow
Train the model using the command:
python training.py model.h5
Run the web application with the trained model:
python app.py model.h5
and then access http://localhost:10000/
