MNIST
MNIST Dataset
The MNIST database of handwritten digits
Dataset Statistics
- Color: Grey-scale
- Sample Size: 28x28
The number of categories of MNIST is 10, that is 0-9, 10 digits.
The Number of Samples per Category for MNIST
| Category | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | Total |
|---|---|---|---|---|---|---|---|---|---|---|---|
| #Training Samples | 5,923 | 6,742 | 5,958 | 6,131 | 5,842 | 5,421 | 5,918 | 6,265 | 5,851 | 5,949 | 60,000 |
| #Testing Samples | 980 | 1,135 | 1,032 | 1,010 | 982 | 892 | 958 | 1,028 | 974 | 1,009 | 10,000 |
Samples
Dataset Usage
MNIST in CSV
The format is:
label, pix-11, pix-12, pix-13, ...
And the script to generate the CSV file from the original dataset is included in this dataset.
Refer to MNIST in CSV
TensorFlow:
TensorFlow provides a simple method for Python to use the MNIST dataset. @tensorflow_MNIST_For_ML_Beginners
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets('MNIST_data/', one_hot=True)
Caffe:
Caffe will download and convert the MNIST dataset to LMDB format throught the scripts. @caffe_Training_LeNet_on_MNIST_with_Caffe
export CAFFE_ROOT='path_to_caffe_root_folder'
cd $CAFFE_ROOT
./data/mnist/get_mnist.sh
./examples/mnist/create_mnist.sh
Torch
Torch will download MNIST automatically by executing:
th train-on-mnist.lua
General tools for Python:
mnist: Python utilities to download and parse the MNIST dataset