MNIST is a dataset commonly used for testing computer vision algorithms. It contains images with handwritten digits. With CAI, MNIST dataset is loaded with:

CreateMNISTVolumes(ImgTrainingVolumes, ImgValidationVolumes, ImgTestVolumes, 'train', 't10k');

It doesn't make sense using FlipX data augmentation method with MNIST. This is disabled with:

NeuralFit.HasFlipX := false;

This simple neural network classifies MNIST with above 99% accuracy as shown in the results folder.