• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1# Copyright 2015 The TensorFlow Authors. All Rights Reserved.
2#
3# Licensed under the Apache License, Version 2.0 (the "License");
4# you may not use this file except in compliance with the License.
5# You may obtain a copy of the License at
6#
7#     http://www.apache.org/licenses/LICENSE-2.0
8#
9# Unless required by applicable law or agreed to in writing, software
10# distributed under the License is distributed on an "AS IS" BASIS,
11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12# See the License for the specific language governing permissions and
13# limitations under the License.
14# ==============================================================================
15"""CIFAR10 small images classification dataset."""
16
17import os
18
19import numpy as np
20
21from tensorflow.python.keras import backend
22from tensorflow.python.keras.datasets.cifar import load_batch
23from tensorflow.python.keras.utils.data_utils import get_file
24from tensorflow.python.util.tf_export import keras_export
25
26
27@keras_export('keras.datasets.cifar10.load_data')
28def load_data():
29  """Loads the CIFAR10 dataset.
30
31  This is a dataset of 50,000 32x32 color training images and 10,000 test
32  images, labeled over 10 categories. See more info at the
33  [CIFAR homepage](https://www.cs.toronto.edu/~kriz/cifar.html).
34
35  The classes are:
36
37  | Label | Description |
38  |:-----:|-------------|
39  |   0   | airplane    |
40  |   1   | automobile  |
41  |   2   | bird        |
42  |   3   | cat         |
43  |   4   | deer        |
44  |   5   | dog         |
45  |   6   | frog        |
46  |   7   | horse       |
47  |   8   | ship        |
48  |   9   | truck       |
49
50  Returns:
51    Tuple of NumPy arrays: `(x_train, y_train), (x_test, y_test)`.
52
53  **x_train**: uint8 NumPy array of grayscale image data with shapes
54    `(50000, 32, 32, 3)`, containing the training data. Pixel values range
55    from 0 to 255.
56
57  **y_train**: uint8 NumPy array of labels (integers in range 0-9)
58    with shape `(50000, 1)` for the training data.
59
60  **x_test**: uint8 NumPy array of grayscale image data with shapes
61    (10000, 32, 32, 3), containing the test data. Pixel values range
62    from 0 to 255.
63
64  **y_test**: uint8 NumPy array of labels (integers in range 0-9)
65    with shape `(10000, 1)` for the test data.
66
67  Example:
68
69  ```python
70  (x_train, y_train), (x_test, y_test) = keras.datasets.cifar10.load_data()
71  assert x_train.shape == (50000, 32, 32, 3)
72  assert x_test.shape == (10000, 32, 32, 3)
73  assert y_train.shape == (50000, 1)
74  assert y_test.shape == (10000, 1)
75  ```
76  """
77  dirname = 'cifar-10-batches-py'
78  origin = 'https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz'
79  path = get_file(
80      dirname,
81      origin=origin,
82      untar=True,
83      file_hash=
84      '6d958be074577803d12ecdefd02955f39262c83c16fe9348329d7fe0b5c001ce')
85
86  num_train_samples = 50000
87
88  x_train = np.empty((num_train_samples, 3, 32, 32), dtype='uint8')
89  y_train = np.empty((num_train_samples,), dtype='uint8')
90
91  for i in range(1, 6):
92    fpath = os.path.join(path, 'data_batch_' + str(i))
93    (x_train[(i - 1) * 10000:i * 10000, :, :, :],
94     y_train[(i - 1) * 10000:i * 10000]) = load_batch(fpath)
95
96  fpath = os.path.join(path, 'test_batch')
97  x_test, y_test = load_batch(fpath)
98
99  y_train = np.reshape(y_train, (len(y_train), 1))
100  y_test = np.reshape(y_test, (len(y_test), 1))
101
102  if backend.image_data_format() == 'channels_last':
103    x_train = x_train.transpose(0, 2, 3, 1)
104    x_test = x_test.transpose(0, 2, 3, 1)
105
106  x_test = x_test.astype(x_train.dtype)
107  y_test = y_test.astype(y_train.dtype)
108
109  return (x_train, y_train), (x_test, y_test)
110