README.md

MNIST-txt

The MNIST training and testing datasets as plain UTF-8 text files.

Formatting

Each line contains comma-separated values representing the image label (0 -9) followed by the pixel's grey values (0 - 255).

More precisely, a line contains 1 label value, 784-pixel values, 784 commas, and one line break.

The pixel values are organized in the same way as the raw IDX files (i.e. row-wise; top to bottom...)

Label,pixel(1,1),pixel(1,2), . . . ,pixel(24,24)