mnist-py 0.5

Lighweight Numpy MNIST loader

mnist-async

Loads from /tmp/mnist/ after downloading any missing files.

Installation

pip install mnist

Usage

from mnist import accuracy, load_mnist, minibatches, render


# Load all the data into memory
mnist = load_mnist()

mnist.train.images  # (60000, 784)
mnist.train.labels  # (60000, 10)
mnist.test.images   # (10000, 784)
mnist.test.labels   # (10000, 10)

# Inspect the data visually
render(train_or_test='test')

# Load the train set into memory and yield minibatches from it
for images, labels in minibatches(batch_size=256, n_epochs=1):

    predicted_labels = ...

    # Calculate the accuracy of predicted class labels
    acc = accuracy(predicted_labels, labels)

Image data

Images are rows, each of length 784, and with pixel values scaled to the range zero through one.

Label data

Lables are one-hot rows each of length ten.

[0 0 1 ... 0]  # 3
[0 0 0 ... 1]  # 9