mindspore.dataset.vision.MixUp

View Source On Gitee
class mindspore.dataset.vision.MixUp(batch_size, alpha, is_single=True)[source]

Randomly mix up a batch of numpy.ndarray images together with its labels.

Each image will be multiplied by a random weight \(lambda\) generated from the Beta distribution and then added to another image multiplied by \(1 - lambda\). The same transformation will be applied to their labels with the same value of \(lambda\). Make sure that the labels are one-hot encoded in advance.

Parameters

  • batch_size (int) – The number of images in a batch.

  • alpha (float) – The alpha and beta parameter for the Beta distribution.

  • is_single (bool, optional) – If True, it will randomly mix up [img0, …, img(n-1), img(n)] with [img1, …, img(n), img0] in each batch. Otherwise, it will randomly mix up images with the output of the previous batch. Default: True.

Raises

  • TypeError – If batch_size is not of type integer.

  • TypeError – If alpha is not of type float.

  • TypeError – If is_single is not of type boolean.

  • ValueError – If batch_size is not positive.

  • ValueError – If alpha is not positive.

Supported Platforms:

CPU

Examples

>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> import mindspore.dataset.transforms as transforms
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(64, 64, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> numpy_slices_dataset = numpy_slices_dataset.map(
...     operations=lambda img: (data, np.random.randint(0, 5, (3, 1))),
...     input_columns=["image"],
...     output_columns=["image", "label"])
>>> # ont hot decode the label
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms.OneHot(10), input_columns="label")
>>> # batch the samples
>>> numpy_slices_dataset = numpy_slices_dataset.batch(batch_size=4)
>>> # finally mix up the images and labels
>>> numpy_slices_dataset = numpy_slices_dataset.map(
...     operations=vision.MixUp(batch_size=1, alpha=0.2),
...     input_columns=["image", "label"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
...     print(item["image"].shape, item["image"].dtype)
...     print(item["label"].shape, item["label"].dtype)
...     break
(4, 64, 64, 3) float64
(4, 3, 10) float64
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> label = np.array([[0, 1]])
>>> output = vision.MixUp(batch_size=2, alpha=0.2, is_single=False)(data, label)
>>> print(output[0].shape, output[0].dtype)
(2, 100, 100, 3) float64
>>> print(output[1].shape, output[1].dtype)
(2, 2) float64
Tutorial Examples: