Handwritten Digit Classifier (ANN)

This project implements a handwritten digit classifier using an Artificial Neural Network trained on the MNIST dataset.

Features

• ANN built using Keras (TensorFlow backend)
• Achieves ~98.35% test accuracy
• Manual image preprocessing (flattening, normalization)
• Custom handwritten digit prediction support

Model Architecture

• Input layer: 784 neurons
• Hidden layer: Dense (ReLU)
• Output layer: 10 neurons (Softmax)

How to Run

1. Install dependencies
   pip install -r requirements.txt
2. Run the Jupyter notebook
3. Test with your own handwritten digit image

Notes

ANNs are sensitive to preprocessing; real handwritten digits require proper centering and normalization to match MNIST distribution.

##Improvements CNN can be used to further enhance the accuracy

##Author Rishii57_