mnist cifar10 fmnist

We evaluate the performance of our proposed protocol using three benchmark datasets. Each dataset is composed of 25 percent of the local data forming the test dataset and 75 percent of the local data forming the training dataset.

MNIST: The dataset is made up of gray pictures that are digits which are written by hand containing ten different classes which provides 60000 training samples in total.

Fashion-MNIST: Different from MNIST which is a dataset that consists of handwritten digits, Fashion-MNIST is a dataset that contains 10 different classes of images that are mainly fashion clothing products.

CIFAR-10: Different from the datasets above, CIFAR-10 is a dataset that has a total of 60000 color images that can be divided into 10 categories. Each sample is a 32*32 color image.