mindspore.dataset.vision.AutoAugment

class mindspore.dataset.vision.AutoAugment(policy=AutoAugmentPolicy.IMAGENET, interpolation=Inter.NEAREST, fill_value=0)

Apply AutoAugment data augmentation method based on AutoAugment: Learning Augmentation Strategies from Data . This operation works only with 3-channel RGB images.

Parameters

• policy (AutoAugmentPolicy, optional) –

  AutoAugment policies learned on different datasets. Default: AutoAugmentPolicy.IMAGENET. It can be AutoAugmentPolicy.IMAGENET, AutoAugmentPolicy.CIFAR10, AutoAugmentPolicy.SVHN. Randomly apply 2 operations from a candidate set. See auto augmentation details in AutoAugmentPolicy.

  • AutoAugmentPolicy.IMAGENET, means to apply AutoAugment learned on ImageNet dataset.

  • AutoAugmentPolicy.CIFAR10, means to apply AutoAugment learned on Cifar10 dataset.

  • AutoAugmentPolicy.SVHN, means to apply AutoAugment learned on SVHN dataset.

• interpolation (Inter, optional) – Image interpolation method defined by Inter . Default: Inter.NEAREST.

• fill_value (Union[int, tuple[int]], optional) – Pixel fill value for the area outside the transformed image. It can be an int or a 3-tuple. If it is a 3-tuple, it is used to fill R, G, B channels respectively. If it is an integer, it is used for all RGB channels. The fill_value values must be in range [0, 255]. Default: 0.

Raises

• TypeError – If policy is not of type mindspore.dataset.vision.AutoAugmentPolicy .

• TypeError – If interpolation is not of type Inter .

• TypeError – If fill_value is not an integer or a tuple of length 3.

• RuntimeError – If given tensor shape is not <H, W, C>.

Supported Platforms:

CPU

Examples

>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import AutoAugmentPolicy, Inter
>>>
>>> transforms_list = [vision.Decode(), vision.AutoAugment(policy=AutoAugmentPolicy.IMAGENET,
...                                                        interpolation=Inter.NEAREST,
...                                                        fill_value=0)]
>>>
>>> image_folder_dataset = ds.ImageFolderDataset("/path/to/image_folder_dataset_directory")
>>> image_folder_dataset = image_folder_dataset.map(operations=transforms_list,
...                                                 input_columns=["image"])

Tutorial Examples:

• Illustration of vision transforms