Module: Chainer::Datasets::CIFAR

Defined in:
lib/chainer/datasets/cifar.rb

Class Method Summary collapse

Class Method Details

.get_cifar(n_classes, with_label, ndim, scale, device: Chainer::Device.default) ⇒ Object

[View source]

14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
 
# File 'lib/chainer/datasets/cifar.rb', line 14

def self.get_cifar(n_classes, with_label, ndim, scale, device: Chainer::Device.default)
  train_table = ::Datasets::CIFAR.new(n_classes: n_classes, type: :train).to_table
  test_table = ::Datasets::CIFAR.new(n_classes: n_classes, type: :test).to_table

  train_data = train_table[:pixels]
  test_data = test_table[:pixels]
  if n_classes == 10
    train_labels = train_table[:label]
    test_labels = test_table[:label]
  else
    train_labels = train_table[:fine_label]
    test_labels = test_table[:fine_label]
  end

  xm = device.xm
  [
    preprocess_cifar(xm::UInt8[*train_data], xm::UInt8[*train_labels], with_label, ndim, scale),
    preprocess_cifar(xm::UInt8[*test_data], xm::UInt8[*test_labels], with_label, ndim, scale)
  ]
end

.get_cifar10(with_label: true, ndim: 3, scale: 1.0) ⇒ Object

[View source]

6
7
8
 
# File 'lib/chainer/datasets/cifar.rb', line 6

def self.get_cifar10(with_label: true, ndim: 3, scale: 1.0)
  get_cifar(10, with_label, ndim, scale)
end

.get_cifar100(with_label: true, ndim: 3, scale: 1.0) ⇒ Object

[View source]

10
11
12
 
# File 'lib/chainer/datasets/cifar.rb', line 10

def self.get_cifar100(with_label: true, ndim: 3, scale: 1.0)
  get_cifar(100, with_label, ndim, scale)
end

.preprocess_cifar(images, labels, withlabel, ndim, scale, device: Chainer::Device.default) ⇒ Object

[View source]

35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
 
# File 'lib/chainer/datasets/cifar.rb', line 35

def self.preprocess_cifar(images, labels, withlabel, ndim, scale, device: Chainer::Device.default)
  if ndim == 1
    images = images.reshape(images.shape[0], 3072)
  elsif ndim == 3
    images = images.reshape(images.shape[0], 3, 32, 32)
  else
    raise 'invalid ndim for CIFAR dataset'
  end
  xm = device.xm
  images = images.cast_to(xm::SFloat)
  images *= scale / 255.0

  if withlabel
    labels = labels.cast_to(xm::Int32)
    TupleDataset.new(images, labels)
  else
    images
  end
end