Custom made Xtrain and Ytrain for neural network training?

Just started reading and learning about neural networks in general and have made a dataset of captcha images to train a neural network on (as a sort of training) the solutions of which are the names of said images. During the reading of documentation I saw that trainNetwork() function takes in training data in all sorts of different ways, and one I am interested in was Xtrain and Ytrain data which X is: "Images, specified as a 4-D numeric array".

In example code they usually just show how to extract X and Y from imds datasets but is there any of examples of custom made Xs and Ys?