MNIST Database of Handwritten Digits

External

Linked on 10/16/2021

Well-known database of 70,000 handwritten digits (10 class labels) with each example represented as an image of 28 x 28 gray-scale pixels.

Dataset Characteristics

Image

Subject Area

Other

Associated Tasks

Classification

Feature Type

Real

# Instances

70000

# Features

-

Dataset Information

For what purpose was the dataset created?

As a testbed for development of handwriting recognition algorithms and machine learning classification algorithms in general.

Who funded the creation of the dataset?

The US National Institute of Standards and Technology (NIST) originally, and later, AT&T Bell Labs

What do the instances in this dataset represent?

28 x 28 gray-scale centered images of handwritten digites

Are there recommended data splits?

Yes. A specific split with 60,000 for training, 10,000 for testing.

Was there any data preprocessing performed?

The original NIST data was preprocessed by Yann LeCun and colleagues at AT&T Bell Labs: see http://yann.lecun.com/exdb/mnist/ for details

Has Missing Values?

No

Introductory Paper

Gradient-based learning applied to document recognition

By Y. LeCun, L. Bottou, Y. Bengio, P. Haffner. 1998

Published in Proceedings of the IEEE

Reviews

There are no reviews for this dataset yet.

Login to Write a Review
Dataset Home Page
1 citations
12131 views

Citations/Acknowledgements

If you use this dataset, please follow the acknowledgment policy on the original dataset website.

Notes

License

By using the UCI Machine Learning Repository, you acknowledge and accept the cookies and privacy practices used by the UCI Machine Learning Repository.

Read Policy