A simple R package to download, collate, and load the classic MNIST dataset in R.

You can install mnistR from GitHub with:

# install.packages("devtools")
devtools::install_github("itsrainingdata/mnistR")

Access the data in one line, as quickly as possible:

MNIST <- mnistR::autoloadMNIST()

By default, the data will be saved to the current working directory. This directory can be manually specified via the dest argument (see Overview).

Download the data to your local drive at /path/to:

mnistR::downloadMNIST(dest = "/path/to/")

Load the data (after downloading):

mnistR::loadMNIST(dest = "/path/to/")

The result is a list containing two data frames (train and test). The labels are stored in the last column, train$y and test$y.

The autoloadMNIST method is simply a wrapper to combine these steps into one method. For even finer grained access, see the methods load_mnist_train, load_mnist_test, and the corresponding download and autoload methods.

This package is a port of code written by @brendano and @daviddalpiaz. The original code can be found here:

• https://gist.github.com/daviddalpiaz/ae62ae5ccd0bada4b9acd6dbc9008706
• https://gist.github.com/brendano/39760