Et astra - a curious collective for inquisitive minds

Ad astra per aspera

itur ad astra

She sold seashells by the seashore
Sphinx of black onyx, hear my plea.

The MNIST database is a large database of handwritten digits that is commonly used for training various image processing systems, and is a seminal dataset in Machine Learning

The below data are the test and training data for the MNIST dataset.
This is the training images for MNIST
This is the training labels for MNIST
This is the test images for MNIST
This is the test labels for MNIST
This is a link to the MNIST page

This is a link to the MNIST Viewer page

Fundamental characteristics of the MNIST database:

1. The digits are number 0-9
2. There are 60,000 images in the training set
3. There are 10,000 images in the test set
4. The digits are 28x28 pixels and stored as arrays instead of
   • JPEG
   • BMP
   • TIF

This differs from other common datasets used in Machine Learning:

ImageNet

Comes in either 1k (more widely used) or 22k varieties. This is the gold standard for image tasks

CIFAR-10

This is another common dataset of classes of 10 items from the Canadians. Wily Canadians!