To showcase the CNN architecture, we will use a classic classification problem: the Modified National Institute of Standards and Technology (MNIST) dataset. The MNIST database is a set of handwritten digits composed of 60,000 datapoints for testing and 10,000 for training.
Keras provides ready to use datasets, including the MNIST dataset. The resolution of the images is only 28 x 28 pixels and is black and white, therefore, the computational cost to train the network is relatively low.
We will now illustrate how to create a CNN in Keras to solve a classification task using the MNIST dataset and achieve human-like performance.
