Skip to content

Latest commit

 

History

History
36 lines (23 loc) · 1.12 KB

README.md

File metadata and controls

36 lines (23 loc) · 1.12 KB

MNIST Digit Classifier

This project demonstrates how to classify handwritten digits using a neural network trained on the MNIST dataset.

Features

  • Preprocesses images for input into the neural network.
  • Trains a simple feedforward neural network using TensorFlow/Keras.
  • Supports testing with custom handwritten digit images.

Installation

  1. Clone the repository: 
    git clone https://github.com/ShouqSaadRu/mnist-digit-classifier.git
cd mnist-digit-classifier
  2. Install dependencies: pip install tensorflow numpy pillow matplotlib

Testing with a Handwritten Image:

I tested the model with this custom handwritten image:

Handwritten Digit

The model correctly predicted the digit as 5. This shows the neural network works well with user-provided input after proper preprocessing.

How to Test Your Own Image: Save your handwritten digit as 28x28 pixels, black digit on a white background. Place the image in the project folder and run this code:

custom_image = preprocess_custom_image("your_image.png") prediction = model.predict(custom_image) print(f"Predicted digit: {np.argmax(prediction)}")