CoolFace
Datasetpublic

teleprint-me/mnist

MNIST Description Raw MNIST dataset as PNG images. Download wget 'https://huggingface.co/datasets/teleprint-me/mnist/resolve/main/mnist.tar.gz?download=true' -O mnist.tar.gz wget 'https://huggingface.co/datasets/teleprint-me/mnist/resolve/main/sha256sum.txt?download=true' -O sha256sum.txt Validate sha256sum -c sha256sum.txt Expected output: mnist.tar.gz: OK Extract tar xf mnist.tar.gz This will create a top-level… See the full description on the dataset page: https://huggingface.co/datasets/teleprint-me/mnist.

sourceHugging Facemitupdated 1y agoView on Hugging Face
0likes15downloads
Dataset Card

MNIST

Description

Raw MNIST dataset as PNG images.

Download

sh
wget 'https://huggingface.co/datasets/teleprint-me/mnist/resolve/main/mnist.tar.gz?download=true' -O mnist.tar.gz
wget 'https://huggingface.co/datasets/teleprint-me/mnist/resolve/main/sha256sum.txt?download=true' -O sha256sum.txt

Validate

sh
sha256sum -c sha256sum.txt

Expected output:

sh
mnist.tar.gz: OK

Extract

sh
tar xf mnist.tar.gz

This will create a top-level mnist/ directory containing the dataset.

Metadata

  • 70,000 images
  • 10 classes (digits 0–9)
  • PNG format
  • 28×28 pixels (784 total pixels per image)
  • 8-bit grayscale
  • Non-interlaced

Disk Usage

sh
du -hs mnist
276M    mnist

Count Total Images

sh
find mnist -type f | wc -l
70000

Count Images per Class

sh
ls -1 mnist/training/0 | wc -l
5923

Inspect Metadata

sh
file mnist/training/0/10005.png
mnist/training/0/10005.png: PNG image data, 28 x 28, 8-bit grayscale, non-interlaced

ImageMagick

sh
identify mnist/training/0/10005.png
mnist/training/0/10005.png PNG 28x28 28x28+0+0 8-bit Grayscale Gray 256c 242B 0.000u 0:00.000

To view detailed metadata, use the -verbose flag:

sh
identify -verbose mnist/training/0/10005.png

Example output:

Image:
  Filename: mnist/training/0/10005.png
  Format: PNG (Portable Network Graphics)
  Geometry: 28x28+0+0
  Colorspace: Gray
  Type: Grayscale
  Depth: 8-bit
  Pixels: 784
  ...

The full output includes histogram, colormap, transparency, and statistical details, which are omitted here for brevity.