Exploring Alternative CNN for Digit and Letter Recognition

Convolutional Neural Networks (CNN) are commonplace for image recognition and are already proven to have high accuracy on datasets such as MNIST and EMNIST. We want to explore alternative methods to improve accuracy on a portion of the EMNIST dataset through the use of multiple CNNs and by using broader data in the initial steps of our CNNs that later in the process will become more specific. It can be seen that by using this method percentages decrease.