* This dataset supports online use.Click here to jump.

MNIST Dataset is a handwritten digit dataset that contains 60,000 example training sets and 10,000 example test sets. It is mainly used for image classification in the field of machine vision. The digits in the dataset have been size-normalized and centered on images of fixed size.

The MNIST Dataset was released by the Courant Institute of New York University, Google New York Laboratory and Microsoft Redmond Research Institute in November 1998. The main publishers were Y. LeCun, L. Bottou, Y. Bengio and P. Haffner. It is currently the most classic training dataset for image classification.