I was learning the basics of TensorFlow and Keras, coding along with a tutorial.
Here TensorFlow MNIST dataset is used which contains 60,000 greyscale 28x28 images in the training set and 10,000 greyscale 28x28 images in the test set. First two images of the dataset are shown in the notebook for preview.
A neural network model is created and it is trained 5 times. By the end of the fifth iteration on the training dataset, the model achieves 98.65% accuracy on determining the hand written number.
To get a realistic view, the model is shown images from the test dataset for classification. It achieves a 97.85% test accuracy.
This means that the model can recognize hand written digits with 97.85% accuracy.