DigitRecognitionML

This project classifies handwritten digits using machine learning models. Models were trained on the MNIST dataset and tested on other datasets then a model was trained on an merged dataset. Detailed information on model training, testing, and evaluation can be found in the Project Report (PDF).

MNIST: Used for training and testing.

EMNIST, USPS, SVHN: Used for testing generalization.

MNIST: https://www.kaggle.com/datasets/hojjatk/mnist-dataset

EMNIST: https://www.kaggle.com/datasets/crawford/emnist

USPS: https://www.kaggle.com/datasets/bistaumanga/usps-dataset

SVHN: http://ufldl.stanford.edu/housenumbers/

A web app built using Streamlit allows users to upload or draw digits for prediction. The K-Nearest Neighbors (KNN) model, which performed the best and trained with a merged datasetis used for predictions.

Downlaod required datasets (EMNIST in csv format and SVHN, MNIST and USPS download is coded)

Install the required dependencies using requirements.txt.

Run "DigitClassifier.ipynb" for model training (if models have not been trained yet).

Run "WebApp.ipynb" (if first time executing it).

Run the Streamlit app: