Handwritten Digit Recognition

Just another repository using CNN to classify handwritten digits

Overview

This repository contains three increasingly improved CNN models for handwritten digit recogition.

• The data used to train model_01 and model_02 is not categoricalized
• model_03 is trained on categoricalized data

Usage

• model_01.py to model_03.py compile and train the respective model, creating a .kreas file
• test_sparse.py model_XX.keras can be used to test the accuracy and loss for models trained on not categoricalized data
• test_categorical.py model_XX.keras can be used to test the accuracy and loss for models trained on categoricalized data
• predicy.py model_XX.keras predicts the number written on images in the ./img folder while visualizing the png