mindspore.dataset.vision.py_transforms.Cutout

class mindspore.dataset.vision.py_transforms.Cutout(length, num_patches=1)[source]

Randomly cut out a certain number of square patches on the input numpy.ndarray image, setting the pixel values in the patch to zero.

See Improved Regularization of Convolutional Neural Networks with Cutout.

Parameters

  • length (int) – The side length of square patches to be cut out.

  • num_patches (int, optional) – The number of patches to be cut out. Default: 1.

Raises

  • TypeError – If length is not of type int.

  • TypeError – If num_patches is not of type int.

  • ValueError – If length is less than or equal 0.

  • ValueError – If num_patches is less than or equal 0.

  • RuntimeError – If shape of the input image is not <H, W, C>.

Supported Platforms:

CPU

Examples

>>> from mindspore.dataset.transforms.py_transforms import Compose
>>>
>>> transforms_list = Compose([py_vision.Decode(),
...                            py_vision.ToTensor(),
...                            py_vision.Cutout(80)])
>>> # apply the transform to dataset through map function
>>> image_folder_dataset = image_folder_dataset.map(operations=transforms_list,
...                                                 input_columns="image")