MLDatasets MNIST Indexing

Why is it that if I look at an image from the MNIST dataset in MLDatasets as a matrix the rows and columns are switched?

using MLDatasets
using ImageShow

trainset= MNIST(:train)

A = trainset.features[:,:,1]

So A[13,2] would actually be the pixel corresponding the 2nd row and the 13th column when I view A as an image.