All libraries we introduced in the last chapter provide support for convolutional layers. We are going to illustrate the LeNet5 architecture using the most basic MNIST handwritten digit dataset, and then use AlexNet on CIFAR10, a simplified version of the original ImageNet, to demonstrate the use of data augmentation.
How to design and train a CNN using Python
LeNet5 and MNIST using Keras
The original MNIST dataset contains 60,000 images in 28 x 28 pixel resolution, with a single grayscale containing handwritten digits from 0 to 9. A good alternative is the more challenging, but structurally similar, Fashion MNIST dataset, which we encountered in Chapter 12, Unsupervised Learning. See the mnist_with_ffnn_and_lenet5 notebook...
