Fashion MNIST Dataset

Estimated reading: 3 minutes 247 views

Visualization of the Fashion MNIST Dataset on the Deep Lake UI

Fashion MNIST Dataset

What is Fashion MNIST Dataset?

The Fashion MNIST (Fashion Modified National Institute of Standards and Technology database) dataset is comprised of 60,000 samples of the training set and 10,000 samples of the test set. Each sample is a 28×28 grayscale picture with a label from one of ten classes. Fashion-MNIST is intended to be a direct drop-in replacement for the original MNIST dataset for evaluating machine learning algorithms. The image size and structure of the training and testing splits are the same.

Download Fashion MNIST Dataset in Python

Instead of downloading the Fashion-MNIST dataset in Python, you can effortlessly load it in Python via our Deep Lake open-source  with just one line of code.

Load Fashion MNIST Dataset Training Subset in Python

					import deeplake
ds = deeplake.load('hub://activeloop/fashion-mnist-train')

Load Fashion MNIST Dataset Testing Subset in Python

					import deeplake
ds = deeplake.load('hub://activeloop/fashion-mnist-test')

Fashion MNIST Dataset Structure

Fashion MNIST Data Fields
  • images: tensor containing the 28×28 image (T-shirt/top, Trouser, Pullover, Dress, Coat, Sandal, Shirt, Sneaker, Bag, Ankle boot).
  • labels: a numerical label that represents the index of the article in the label list.
Fashion MNIST Data Splits
  • The Fashion MNIST dataset training set is composed of 60,000 examples.
  • The Fashion MNIST dataset test set was composed of 10,000 examples.

How to use Fashion MNIST Dataset with PyTorch and TensorFlow in Python

Train a model on Fashion MNIST dataset with PyTorch in Python

Let’s use Deep Lake built-in PyTorch one-line dataloader to connect the data to the compute:

					dataloader = ds.pytorch(num_workers=0, batch_size=4, shuffle=False)
Train a model on Fashion MNIST dataset with TensorFlow in Python
					dataloader = ds.tensorflow()

Additional Information about Fashion MNIST Dataset

Fashion MNIST Dataset Description

Fashion MNIST Dataset Curators
Han Xiao, Kashif Rasul, Roland Vollgraf
Fashion MNIST Dataset Licensing Information
MIT Licence
Fashion MNIST Dataset Citation Information
title={Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms},
author={Xiao, Han and Rasul, Kashif and Vollgraf, Roland},
journal={arXiv preprint arXiv:1708.07747},

Fashion MNIST Dataset FAQs

What are the classes of images in the Fashion MNIST dataset?
Similar to MNIST, Fashion-MNIST contains 10 classes of images.
  • T-shirt/top,
  • Trouser,
  • Pullover,
  • Dress,
  • Coat,
  • Sandal,
  • Shirt,
  • Sneaker,
  • Bag,
  • Ankle boot.
What are the main differences between MNIST and Fashion MNIST dataset?
  • Fashion MNIST is meant to substitute the MNIST dataset. Both datasets have the same amount of pictures in the training set (60 000 pictures) as well as the testing set (10 000 pictures).
  • Both Fashion MNIST and MNIST datasets have 10 classes: the ten digits (0 to 9) for MNIST, whilst Fashion MNIST is ten kinds of clothing items.
  • Both datasets consist of 28×28 pixel greyscale pictures, with each pixel being a number between 0 and 255 representing the greyscale intensity.
How many images are in the Fashion MNIST dataset?
  • The Fashion MNIST dataset has 70 000 images, consisting of 60 000 training set and 10 000 testing set images.