Hand-Written Digit Recognizer

Digit recognition on hand-written digits using Convolutional Neural Network with Keras library. Achieved an accuracy of 99.50%.

Setup :-

Python --> Python 3.6.5

OS --> Ubuntu 19.04 / Windows 10

GPU --> Nvidia Geforce 940MX (2GB)

Intel Core i5-7200 @ 2.50GHz

RAM --> 4gb

Using mnist_train2.py :-

Model can be trained by running mnist_train2.py

python mnist_train2.py

Trains the model to predict hand-written digits from their images.

Saves the weights in model.h5 file.

Saves the model in model.json file.

Using mnist_test2.py :-

Model can be tested by running mnist_test2.py

python mnist_test2.py

Predicts hand-written digits from their images.

Loads the weights from model.h5 file.

Loads the model from model.json file.

Displays a random image of a hand-written digit on screen.

Predicts the digit and prints it on terminal.