Permuted MNIST Digit Image Dataset

Permuted MNIST is a pixel-permuted version of the MNIST dataset, which consists of 70,000 images of handwritten digits from 0 to 9, of which 60,000 images are used for training and 10,000 images are used for testing.

The difference between this dataset and the original MNIST dataset is that each task is to perform multiple classifications on different random arrangements of the input pixels.