Skip to content

Files

Latest commit

alexbuckgitalexbuckgit
[BULK UPDATE] DocuTune - Apply event-tier1-build-2022 ms.custom tag t…
May 23, 2022
55762f7 · May 23, 2022

History

History
330 lines (256 loc) · 10.8 KB

dataset-mnist.md

File metadata and controls

330 lines (256 loc) · 10.8 KB
title description ms.service ms.custom ms.topic ms.date
MNIST database of handwritten digits
Learn how to use the MNIST database of handwritten digits dataset in Azure Open Datasets.
open-datasets
event-tier1-build-2022
sample
04/16/2021

The MNIST database of handwritten digits

The MNIST database of handwritten digits has a training set of 60,000 examples and a test set of 10,000 examples. The digits have been size-normalized and centered in a fixed-size image.

[!INCLUDE Open Dataset usage notice]

This dataset is sourced from THE MNIST DATABASE of handwritten digits. It's a subset of the larger NIST Hand-printed Forms and Characters Database published by National Institute of Standards and Technology.

Storage location

  • Blob account: azureopendatastorage
  • Container name: mnist

Four files are available in the container directly:

  • train-images-idx3-ubyte.gz: training set images (9,912,422 bytes)
  • train-labels-idx1-ubyte.gz: training set labels (28,881 bytes)
  • t10k-images-idx3-ubyte.gz: test set images (1,648,877 bytes)
  • t10k-labels-idx1-ubyte.gz: test set labels (4,542 bytes)

Data access

Azure Notebooks

azureml-opendatasets

Tip

Download the notebook instead.

Load MNIST into a data frame using Azure Machine Learning tabular datasets.

For more information on Azure Machine Learning datasets, see Create Azure Machine Learning datasets.

Get complete dataset into a data frame

from azureml.opendatasets import MNIST

mnist = MNIST.get_tabular_dataset()
mnist_df = mnist.to_pandas_dataframe()
mnist_df.info()

Get train and test data frames

mnist_train = MNIST.get_tabular_dataset(dataset_filter='train')
mnist_train_df = mnist_train.to_pandas_dataframe()
X_train = mnist_train_df.drop("label", axis=1).astype(int).values/255.0
y_train = mnist_train_df.filter(items=["label"]).astype(int).values

mnist_test = MNIST.get_tabular_dataset(dataset_filter='test')
mnist_test_df = mnist_test.to_pandas_dataframe()
X_test = mnist_test_df.drop("label", axis=1).astype(int).values/255.0
y_test = mnist_test_df.filter(items=["label"]).astype(int).values

Plot some images of the digits

%matplotlib inline
import numpy as np
import matplotlib.pyplot as plt

# now let's show some randomly chosen images from the traininng set.
count = 0
sample_size = 30
plt.figure(figsize=(16, 6))
for i in np.random.permutation(X_train.shape[0])[:sample_size]:
    count = count + 1
    plt.subplot(1, sample_size, count)
    plt.axhline('')
    plt.axvline('')
    plt.text(x=10, y=-10, s=y_train[i], fontsize=18)
    plt.imshow(X_train[i].reshape(28, 28), cmap=plt.cm.Greys)
plt.show()

Download or mount MNIST raw files Azure Machine Learning file datasets.

This works only for Linux based compute. For more information on Azure Machine Learning datasets, see Create Azure Machine Learning datasets.

mnist_file = MNIST.get_file_dataset()
mnist_file
mnist_file.to_path()

Download files to local storage

import os
import tempfile

data_folder = tempfile.mkdtemp()
data_paths = mnist_file.download(data_folder, overwrite=True)
data_paths

Mount files. Useful when training job will run on a remote compute.

import gzip
import struct
import pandas as pd
import numpy as np

# load compressed MNIST gz files and return pandas dataframe of numpy arrays
def load_data(filename, label=False):
    with gzip.open(filename) as gz:
        gz.read(4)
        n_items = struct.unpack('>I', gz.read(4))
        if not label:
            n_rows = struct.unpack('>I', gz.read(4))[0]
            n_cols = struct.unpack('>I', gz.read(4))[0]
            res = np.frombuffer(gz.read(n_items[0] * n_rows * n_cols), dtype=np.uint8)
            res = res.reshape(n_items[0], n_rows * n_cols)
        else:
            res = np.frombuffer(gz.read(n_items[0]), dtype=np.uint8)
            res = res.reshape(n_items[0], 1)
    return pd.DataFrame(res)
import sys
mount_point = tempfile.mkdtemp()
print(mount_point)
print(os.path.exists(mount_point))

if sys.platform == 'linux':
  print("start mounting....")
  with mnist_file.mount(mount_point):
    print("list dir...")
    print(os.listdir(mount_point))
    print("get the dataframe info of mounted data...")
    train_images_df = load_data(next(path for path in data_paths if path.endswith("train-images-idx3-ubyte.gz")))
    print(train_images_df.info())

azure-storage

Tip

Download the notebook instead.

%matplotlib inline
import numpy as np
import matplotlib.pyplot as plt
import urllib.request
import os

data_folder = os.path.join(os.getcwd(), 'data')
os.makedirs(data_folder, exist_ok=True)

urllib.request.urlretrieve('https://azureopendatastorage.blob.core.windows.net/mnist/train-images-idx3-ubyte.gz',
                           filename=os.path.join(data_folder, 'train-images.gz'))
urllib.request.urlretrieve('https://azureopendatastorage.blob.core.windows.net/mnist/train-labels-idx1-ubyte.gz',
                           filename=os.path.join(data_folder, 'train-labels.gz'))
urllib.request.urlretrieve('https://azureopendatastorage.blob.core.windows.net/mnist/t10k-images-idx3-ubyte.gz',
                           filename=os.path.join(data_folder, 'test-images.gz'))
urllib.request.urlretrieve('https://azureopendatastorage.blob.core.windows.net/mnist/t10k-labels-idx1-ubyte.gz',
                           filename=os.path.join(data_folder, 'test-labels.gz'))
import gzip
import struct

# load compressed MNIST gz files and return numpy arrays
def load_data(filename, label=False):
    with gzip.open(filename) as gz:
        struct.unpack('I', gz.read(4))
        n_items = struct.unpack('>I', gz.read(4))
        if not label:
            n_rows = struct.unpack('>I', gz.read(4))[0]
            n_cols = struct.unpack('>I', gz.read(4))[0]
            res = np.frombuffer(gz.read(n_items[0] * n_rows * n_cols), dtype=np.uint8)
            res = res.reshape(n_items[0], n_rows * n_cols)
        else:
            res = np.frombuffer(gz.read(n_items[0]), dtype=np.uint8)
            res = res.reshape(n_items[0], 1)
    return res
# note we also shrink the intensity values (X) from 0-255 to 0-1. This helps the model converge faster.
X_train = load_data(os.path.join(
    data_folder, 'train-images.gz'), False) / 255.0
X_test = load_data(os.path.join(data_folder, 'test-images.gz'), False) / 255.0
y_train = load_data(os.path.join(
    data_folder, 'train-labels.gz'), True).reshape(-1)
y_test = load_data(os.path.join(
    data_folder, 'test-labels.gz'), True).reshape(-1)

# now let's show some randomly chosen images from the traininng set.
count = 0
sample_size = 30
plt.figure(figsize=(16, 6))
for i in np.random.permutation(X_train.shape[0])[:sample_size]:
    count = count + 1
    plt.subplot(1, sample_size, count)
    plt.axhline('')
    plt.axvline('')
    plt.text(x=10, y=-10, s=y_train[i], fontsize=18)
    plt.imshow(X_train[i].reshape(28, 28), cmap=plt.cm.Greys)
plt.show()

Azure Databricks

azureml-opendatasets

Tip

Download the notebook instead.

Load MNIST into a data frame using Azure Machine Learning tabular datasets.

For more information on Azure Machine Learning datasets, see Create Azure Machine Learning datasets.

Get complete dataset into a data frame

# This is a package in preview.
from azureml.opendatasets import MNIST

mnist = MNIST.get_tabular_dataset()
mnist_df = mnist.to_spark_dataframe()
display(mnist_df.limit(5))

Download or mount MNIST raw files Azure Machine Learning file datasets.

This works only for Linux based compute. For more information on Azure Machine Learning datasets, see Create Azure Machine Learning datasets.

mnist_file = MNIST.get_file_dataset()
mnist_file
mnist_file.to_path()

Download files to local storage

import os
import tempfile

mount_point = tempfile.mkdtemp()
mnist_file.download(mount_point, overwrite=True)

Mount files. Useful when training job will run on a remote compute.

import gzip
import struct
import pandas as pd
import numpy as np

# load compressed MNIST gz files and return numpy arrays
def load_data(filename, label=False):
    with gzip.open(filename) as gz:
        gz.read(4)
        n_items = struct.unpack('>I', gz.read(4))
        if not label:
            n_rows = struct.unpack('>I', gz.read(4))[0]
            n_cols = struct.unpack('>I', gz.read(4))[0]
            res = np.frombuffer(gz.read(n_items[0] * n_rows * n_cols), dtype=np.uint8)
            res = res.reshape(n_items[0], n_rows * n_cols)
        else:
            res = np.frombuffer(gz.read(n_items[0]), dtype=np.uint8)
            res = res.reshape(n_items[0], 1)
    return pd.DataFrame(res)
import sys
mount_point = tempfile.mkdtemp()
print(mount_point)
print(os.path.exists(mount_point))
print(os.listdir(mount_point))

if sys.platform == 'linux':
  print("start mounting....")
  with mnist_file.mount(mount_point):
    print(context.mount_point )
    print(os.listdir(mount_point))  
    train_images_df = load_data(os.path.join(mount_point, 'train-images-idx3-ubyte.gz'))
    print(train_images_df.info())

azure-storage

Sample not available for this platform/package combination.

Next steps

View the rest of the datasets in the Open Datasets catalog.