# Copyright 2019-2022 Huawei Technologies Co., Ltd
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""
The module vision.transforms provides many kinds of image augmentation methods
and image-related conversion methods
(e.g. including with PIL.Image.Image and numpy.ndarray).
to perform various computer vision tasks.
Users can apply suitable augmentations on image data
to improve their training models.
Users can also self-define their own augmentation methods with Python Pillow (PIL)
For the different methods in this module, implementation is based in C++ and/or Python.
The C++ implementation is inherited from mindspore._c_dataengine, provides high performance
and is mainly based on OpenCV.
The Python implementation is mainly based on PIL.
.. Note::
A constructor's arguments for every class in this module must be saved into the
class attributes (self.xxx) to support save() and load().
Examples:
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Border, Inter
>>> import mindspore.dataset.transforms as transforms
>>>
>>> image_folder_dataset_dir = "/path/to/image_folder_dataset_directory"
>>> # create a dataset that reads all files in dataset_dir with 8 threads
>>> image_folder_dataset = ds.ImageFolderDataset(image_folder_dataset_dir,
... num_parallel_workers=8)
>>> # create a list of transformations to be applied to the image data
>>> transforms_list = [vision.Decode(),
... vision.Resize((256, 256), interpolation=Inter.LINEAR),
... vision.RandomCrop(200, padding_mode=Border.EDGE),
... vision.RandomRotation((0, 15)),
... vision.Normalize((100, 115.0, 121.0), (71.0, 68.0, 70.0)),
... vision.HWC2CHW()]
>>> onehot_op = transforms.OneHot(num_classes=10)
>>> # apply the transformation to the dataset through data1.map()
>>> image_folder_dataset = image_folder_dataset.map(operations=transforms_list,
... input_columns="image")
>>> image_folder_dataset = image_folder_dataset.map(operations=onehot_op,
... input_columns="label")
"""
# pylint: disable=too-few-public-methods
import numbers
import random
import numpy as np
from PIL import Image
import mindspore._c_dataengine as cde
from mindspore._c_expression import typing
from . import py_transforms_util as util
from .py_transforms_util import is_pil
from .utils import AutoAugmentPolicy, Border, ConvertMode, ImageBatchFormat, Inter, SliceMode, parse_padding
from .validators import check_adjust_brightness, check_adjust_contrast, check_adjust_gamma, check_adjust_hue, \
check_adjust_saturation, check_adjust_sharpness, check_affine, check_alpha, check_auto_augment, \
check_auto_contrast, check_bounding_box_augment_cpp, check_center_crop, check_convert_color, check_crop, \
check_cut_mix_batch_c, check_cutout_new, check_decode, check_erase, check_five_crop, check_gaussian_blur, \
check_hsv_to_rgb, check_linear_transform, check_mix_up, check_mix_up_batch_c, check_normalize, \
check_normalizepad, check_num_channels, check_pad, check_pad_to_size, check_perspective, check_positive_degrees, \
check_posterize, check_prob, check_rand_augment, check_random_adjust_sharpness, check_random_affine, \
check_random_auto_contrast, check_random_color_adjust, check_random_crop, check_random_erasing, \
check_random_perspective, check_random_posterize, check_random_resize_crop, check_random_rotation, \
check_random_select_subpolicy_op, check_random_solarize, check_range, check_rescale, check_resize, \
check_resize_interpolation, check_resized_crop, check_rgb_to_hsv, check_rotate, check_slice_patches, \
check_solarize, check_ten_crop, check_trivial_augment_wide, check_uniform_augment, check_to_tensor, \
check_device_target, FLOAT_MAX_INTEGER
from ..core.datatypes import mstype_to_detype, nptype_to_detype
from ..transforms.py_transforms_util import Implementation
from ..transforms.transforms import CompoundOperation, PyTensorOperation, TensorOperation, TypeCast
class ImageTensorOperation(TensorOperation):
"""
Base class of Image Tensor Ops.
"""
def __call__(self, *input_tensor_list):
for tensor in input_tensor_list:
if not isinstance(tensor, (np.ndarray, Image.Image)):
raise TypeError(
"Input should be NumPy or PIL image, got {}.".format(type(tensor)))
return super().__call__(*input_tensor_list)
def parse(self):
# Note: subclasses must implement `def parse(self)` so do not make ImageTensorOperation's parse a staticmethod.
raise NotImplementedError("ImageTensorOperation has to implement parse() method.")
[docs]class AdjustBrightness(ImageTensorOperation, PyTensorOperation):
"""
Adjust the brightness of the input image.
Supports Ascend hardware acceleration and can be enabled through the `.device("Ascend")` method.
Args:
brightness_factor (float): How much to adjust the brightness, must be non negative.
``0`` gives a black image, ``1`` gives the original image,
while ``2`` increases the brightness by a factor of 2.
Raises:
TypeError: If `brightness_factor` is not of type float.
ValueError: If `brightness_factor` is less than 0.
RuntimeError: If shape of the input image is not <H, W, C>.
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.AdjustBrightness(brightness_factor=2.0)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 256, (20, 20, 3)) / 255.0
>>> data = data.astype(np.float32)
>>> output = vision.AdjustBrightness(2.666)(data)
>>> print(output.shape, output.dtype)
(20, 20, 3) float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_adjust_brightness
def __init__(self, brightness_factor):
super().__init__()
self.brightness_factor = brightness_factor
[docs] @check_device_target
def device(self, device_target="CPU"):
"""
Set the device for the current operator execution.
- When the device is Ascend, input shape should be limited from [4, 6] to [8192, 4096].
Args:
device_target (str, optional): The operator will be executed on this device. Currently supports
``CPU`` and ``Ascend`` . Default: ``CPU`` .
Raises:
TypeError: If `device_target` is not of type str.
ValueError: If `device_target` is not within the valid set of ['CPU', 'Ascend'].
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.AdjustBrightness(2.0).device("Ascend")]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 256, (20, 20, 3)) / 255.0
>>> data = data.astype(np.float32)
>>> output = vision.AdjustBrightness(2.666).device("Ascend")(data)
>>> print(output.shape, output.dtype)
(20, 20, 3) float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
self.device_target = device_target
return self
def parse(self):
return cde.AdjustBrightnessOperation(self.brightness_factor, self.device_target)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be brightness adjusted.
Returns:
PIL Image, brightness adjusted image.
"""
return util.adjust_brightness(img, self.brightness_factor)
[docs]class AdjustContrast(ImageTensorOperation, PyTensorOperation):
"""
Adjust the contrast of the input image.
Supports Ascend hardware acceleration and can be enabled through the `.device("Ascend")` method.
Args:
contrast_factor (float): How much to adjust the contrast, must be non negative.
``0`` gives a solid gray image, ``1`` gives the original image,
while ``2`` increases the contrast by a factor of 2.
Raises:
TypeError: If `contrast_factor` is not of type float.
ValueError: If `contrast_factor` is less than 0.
RuntimeError: If shape of the input image is not <H, W, C>.
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.AdjustContrast(contrast_factor=2.0)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.AdjustContrast(2.0)(data)
>>> print(output.shape, output.dtype)
(2, 2, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_adjust_contrast
def __init__(self, contrast_factor):
super().__init__()
self.contrast_factor = contrast_factor
[docs] @check_device_target
def device(self, device_target="CPU"):
"""
Set the device for the current operator execution.
- When the device is Ascend, input shape should be limited from [4, 6] to [8192, 4096].
Args:
device_target (str, optional): The operator will be executed on this device. Currently supports
``CPU`` and ``Ascend`` . Default: ``CPU`` .
Raises:
TypeError: If `device_target` is not of type str.
ValueError: If `device_target` is not within the valid set of ['CPU', 'Ascend'].
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.AdjustContrast(0).device("Ascend")]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.AdjustContrast(2.0).device("Ascend")(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
self.device_target = device_target
return self
def parse(self):
return cde.AdjustContrastOperation(self.contrast_factor, self.device_target)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be contrast adjusted.
Returns:
PIL Image, contrast adjusted image.
"""
return util.adjust_contrast(img, self.contrast_factor)
[docs]class AdjustGamma(ImageTensorOperation, PyTensorOperation):
r"""
Apply gamma correction on input image. Input image is expected to be in <..., H, W, C> or <H, W> format.
.. math::
I_{\text{out}} = 255 \times \text{gain} \times \left(\frac{I_{\text{in}}}{255}\right)^{\gamma}
See `Gamma Correction`_ for more details.
.. _Gamma Correction: https://en.wikipedia.org/wiki/Gamma_correction
Args:
gamma (float): Non negative real number.
The output image pixel value is exponentially related to the input image pixel value.
gamma larger than 1 make the shadows darker,
while gamma smaller than 1 make dark regions lighter.
gain (float, optional): The constant multiplier. Default: ``1.0``.
Raises:
TypeError: If `gain` is not of type float.
TypeError: If `gamma` is not of type float.
ValueError: If `gamma` is less than 0.
RuntimeError: If given tensor shape is not <H, W> or <..., H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.AdjustGamma(gamma=10.0, gain=1.0)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.AdjustGamma(gamma=0.1, gain=1.0)(data)
>>> print(output.shape, output.dtype)
(2, 2, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_adjust_gamma
def __init__(self, gamma, gain=1):
super().__init__()
self.gamma = gamma
self.gain = gain
self.random = False
def parse(self):
return cde.AdjustGammaOperation(self.gamma, self.gain)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be gamma adjusted.
Returns:
PIL Image, gamma adjusted image.
"""
return util.adjust_gamma(img, self.gamma, self.gain)
[docs]class AdjustHue(ImageTensorOperation, PyTensorOperation):
"""
Adjust the hue of the input image.
Supports Ascend hardware acceleration and can be enabled through the `.device("Ascend")` method.
Args:
hue_factor (float): How much to add to the hue channel,
must be in range of [-0.5, 0.5].
Raises:
TypeError: If `hue_factor` is not of type float.
ValueError: If `hue_factor` is not in the interval [-0.5, 0.5].
RuntimeError: If shape of the input image is not <H, W, C>.
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.AdjustHue(hue_factor=0.2)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.AdjustHue(hue_factor=0.2)(data)
>>> print(output.shape, output.dtype)
(2, 2, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_adjust_hue
def __init__(self, hue_factor):
super().__init__()
self.hue_factor = hue_factor
[docs] @check_device_target
def device(self, device_target="CPU"):
"""
Set the device for the current operator execution.
- When the device is Ascend, input shape should be limited from [4, 6] to [8192, 4096].
Args:
device_target (str, optional): The operator will be executed on this device. Currently supports
``CPU`` and ``Ascend`` . Default: ``CPU`` .
Raises:
TypeError: If `device_target` is not of type str.
ValueError: If `device_target` is not within the valid set of ['CPU', 'Ascend'].
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.AdjustHue(0.5).device("Ascend")]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.AdjustHue(hue_factor=0.2).device("Ascend")(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
self.device_target = device_target
return self
def parse(self):
return cde.AdjustHueOperation(self.hue_factor, self.device_target)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be hue adjusted.
Returns:
PIL Image, hue adjusted image.
"""
return util.adjust_hue(img, self.hue_factor)
[docs]class AdjustSaturation(ImageTensorOperation, PyTensorOperation):
"""
Adjust the saturation of the input image.
Supports Ascend hardware acceleration and can be enabled through the `.device("Ascend")` method.
Args:
saturation_factor (float): How much to adjust the saturation, must be non negative.
``0`` gives a black image, ``1`` gives the original image
while ``2`` increases the saturation by a factor of 2.
Raises:
TypeError: If `saturation_factor` is not of type float.
ValueError: If `saturation_factor` is less than 0.
RuntimeError: If shape of the input image is not <H, W, C>.
RuntimeError: If channel of the input image is not 3.
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.AdjustSaturation(saturation_factor=2.0)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.AdjustSaturation(saturation_factor=2.0)(data)
>>> print(output.shape, output.dtype)
(2, 2, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_adjust_saturation
def __init__(self, saturation_factor):
super().__init__()
self.saturation_factor = saturation_factor
[docs] @check_device_target
def device(self, device_target="CPU"):
"""
Set the device for the current operator execution.
- When the device is Ascend, input shape should be limited from [4, 6] to [8192, 4096].
Args:
device_target (str, optional): The operator will be executed on this device. Currently supports
``CPU`` and ``Ascend`` . Default: ``CPU`` .
Raises:
TypeError: If `device_target` is not of type str.
ValueError: If `device_target` is not within the valid set of ['CPU', 'Ascend'].
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.AdjustSaturation(2.0).device("Ascend")]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.AdjustSaturation(saturation_factor=2.0).device("Ascend")(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
self.device_target = device_target
return self
def parse(self):
return cde.AdjustSaturationOperation(self.saturation_factor, self.device_target)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be saturation adjusted.
Returns:
PIL Image, saturation adjusted image.
"""
return util.adjust_saturation(img, self.saturation_factor)
[docs]class AdjustSharpness(ImageTensorOperation):
"""
Adjust the sharpness of the input image.
Args:
sharpness_factor (float): How much to adjust the sharpness, must be
non negative. ``0`` gives a blurred image, ``1`` gives the
original image while ``2`` increases the sharpness by a factor of 2.
Raises:
TypeError: If `sharpness_factor` is not of type float.
ValueError: If `sharpness_factor` is less than 0.
RuntimeError: If shape of the input image is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> # create a dataset that reads all files in dataset_dir with 8 threads
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.AdjustSharpness(sharpness_factor=2.0)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((3, 4))
>>> output = vision.AdjustSharpness(sharpness_factor=0)(data)
>>> print(output.shape, output.dtype)
(3, 4) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_adjust_sharpness
def __init__(self, sharpness_factor):
super().__init__()
self.sharpness_factor = sharpness_factor
self.implementation = Implementation.C
def parse(self):
return cde.AdjustSharpnessOperation(self.sharpness_factor)
[docs]class Affine(ImageTensorOperation):
"""
Apply Affine transformation to the input image, keeping the center of the image unchanged.
Supports Ascend hardware acceleration and can be enabled through the `.device("Ascend")` method.
Args:
degrees (float): Rotation angle in degrees between -180 and 180, clockwise direction.
translate (Sequence[float, float]): The horizontal and vertical translations, must be a sequence of size 2
and value between -1 and 1.
scale (float): Scaling factor, which must be positive.
shear (Union[float, Sequence[float, float]]): Shear angle value in degrees between -180 to 180.
If float is provided, shear along the x axis with this value, without shearing along the y axis;
If Sequence[float, float] is provided, shear along the x axis and y axis with these two values separately.
resample (Inter, optional): Image interpolation method defined by :class:`~.vision.Inter` .
Default: ``Inter.NEAREST``.
fill_value (Union[int, tuple[int, int, int]], optional): Optional `fill_value` to fill the area
outside the transform in the output image. There must be three elements in tuple and the value
of single element is [0, 255]. Default: ``0``.
Raises:
TypeError: If `degrees` is not of type float.
TypeError: If `translate` is not of type Sequence[float, float].
TypeError: If `scale` is not of type float.
ValueError: If `scale` is non positive.
TypeError: If `shear` is not of float or Sequence[float, float].
TypeError: If `resample` is not of type :class:`~.vision.Inter` .
TypeError: If `fill_value` is not of type int or tuple[int, int, int].
RuntimeError: If shape of the input image is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> affine_op = vision.Affine(degrees=15, translate=[0.2, 0.2], scale=1.1, shear=[1.0, 1.0],
... resample=Inter.BILINEAR)
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=[affine_op], input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.Affine(degrees=15, translate=[0.2, 0.2], scale=1.1,
... shear=[1.0, 1.0], resample=Inter.BILINEAR)(data)
>>> print(output.shape, output.dtype)
(2, 2, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_affine
def __init__(self, degrees, translate, scale, shear, resample=Inter.NEAREST, fill_value=0):
super().__init__()
# Parameter checking
if isinstance(shear, numbers.Number):
shear = (shear, 0.)
if isinstance(fill_value, numbers.Number):
fill_value = (fill_value, fill_value, fill_value)
self.degrees = degrees
self.translate = translate
self.scale_ = scale
self.shear = shear
self.resample = resample
self.fill_value = fill_value
self.implementation = Implementation.C
[docs] @check_device_target
def device(self, device_target="CPU"):
"""
Set the device for the current operator execution.
- When the device is Ascend, input shape should be limited from [4, 6] to [32768, 32768].
Args:
device_target (str, optional): The operator will be executed on this device. Currently supports
``CPU`` and ``Ascend`` . Default: ``CPU`` .
Raises:
TypeError: If `device_target` is not of type str.
ValueError: If `device_target` is not within the valid set of ['CPU', 'Ascend'].
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> affine_op = vision.Affine(degrees=15, translate=[0.2, 0.2], scale=1.1,
... shear=[1.0, 1.0], resample=Inter.BILINEAR).device("Ascend")
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=[affine_op], input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.Affine(degrees=15, translate=[0.2, 0.2], scale=1.1,
... shear=[1.0, 1.0], resample=Inter.BILINEAR).device("Ascend")(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
self.device_target = device_target
return self
def parse(self):
return cde.AffineOperation(self.degrees, self.translate, self.scale_, self.shear,
Inter.to_c_type(self.resample), self.fill_value, self.device_target)
[docs]class AutoAugment(ImageTensorOperation):
"""
Apply AutoAugment data augmentation method based on
`AutoAugment: Learning Augmentation Strategies from Data <https://arxiv.org/pdf/1805.09501.pdf>`_ .
This operation works only with 3-channel RGB images.
Args:
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 :class:`~.vision.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 :class:`mindspore.dataset.vision.AutoAugmentPolicy` .
TypeError: If `interpolation` is not of type :class:`~.vision.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 numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import AutoAugmentPolicy, Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> transforms_list = [vision.AutoAugment(policy=AutoAugmentPolicy.IMAGENET,
... interpolation=Inter.NEAREST,
... fill_value=0)]
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.AutoAugment()(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_auto_augment
def __init__(self, policy=AutoAugmentPolicy.IMAGENET, interpolation=Inter.NEAREST, fill_value=0):
super().__init__()
self.policy = policy
self.interpolation = interpolation
if isinstance(fill_value, int):
fill_value = tuple([fill_value] * 3)
self.fill_value = fill_value
self.implementation = Implementation.C
def parse(self):
return cde.AutoAugmentOperation(AutoAugmentPolicy.to_c_type(self.policy), Inter.to_c_type(self.interpolation),
self.fill_value)
[docs]class AutoContrast(ImageTensorOperation, PyTensorOperation):
"""
Apply automatic contrast on input image. This operation calculates histogram of image, reassign cutoff percent
of the lightest pixels from histogram to 255, and reassign cutoff percent of the darkest pixels from histogram to 0.
Args:
cutoff (float, optional): Percent of lightest and darkest pixels to cut off from
the histogram of input image. The value must be in the range [0.0, 50.0]. Default: ``0.0``.
ignore (Union[int, sequence], optional): The background pixel values to ignore,
The ignore values must be in range [0, 255]. Default: ``None``.
Raises:
TypeError: If `cutoff` is not of type float.
TypeError: If `ignore` is not of type int or sequence.
ValueError: If `cutoff` is not in range [0, 50.0).
ValueError: If `ignore` is not in range [0, 255].
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.AutoContrast(cutoff=10.0, ignore=[10, 20])]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.AutoContrast(cutoff=10.0, ignore=[10, 20])(data)
>>> print(output.shape, output.dtype)
(2, 2, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_auto_contrast
def __init__(self, cutoff=0.0, ignore=None):
super().__init__()
if ignore is None:
ignore = []
if isinstance(ignore, int):
ignore = [ignore]
self.cutoff = cutoff
self.ignore = ignore
self.random = False
def parse(self):
return cde.AutoContrastOperation(self.cutoff, self.ignore)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be automatically contrasted.
Returns:
PIL Image, automatically contrasted image.
"""
return util.auto_contrast(img, self.cutoff, self.ignore)
[docs]class BoundingBoxAugment(ImageTensorOperation):
"""
Apply a given image processing operation on a random selection of bounding box regions of a given image.
Args:
transform (TensorOperation): Transformation operation to be applied on random selection
of bounding box regions of a given image.
ratio (float, optional): Ratio of bounding boxes to apply augmentation on.
Range: [0.0, 1.0]. Default: ``0.3``.
Raises:
TypeError: If `transform` is an image processing operation in `mindspore.dataset.vision` .
TypeError: If `ratio` is not of type float.
ValueError: If `ratio` is not in range [0.0, 1.0].
RuntimeError: If given bounding box is invalid.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.float32)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> func = lambda img: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(np.float32))
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=[func],
... input_columns=["image"],
... output_columns=["image", "bbox"])
>>> # set bounding box operation with ratio of 1 to apply rotation on all bounding boxes
>>> bbox_aug_op = vision.BoundingBoxAugment(vision.RandomRotation(90), 1)
>>> # map to apply ops
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=[bbox_aug_op],
... input_columns=["image", "bbox"],
... output_columns=["image", "bbox"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... print(item["bbox"].shape, item["bbox"].dtype)
... break
(100, 100, 3) float32
(1, 4) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((3, 4))
>>> data = data.astype(np.float32)
>>> func = lambda img, bboxes: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(bboxes.dtype))
>>> func_data, func_bboxes = func(data, data)
>>> output = vision.BoundingBoxAugment(transforms.Fill(100), 1.0)(func_data, func_bboxes)
>>> print(output[0].shape, output[0].dtype)
(3, 4) float32
>>> print(output[1].shape, output[1].dtype)
(1, 4) float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_bounding_box_augment_cpp
def __init__(self, transform, ratio=0.3):
super().__init__()
self.ratio = ratio
self.transform = transform
self.implementation = Implementation.C
def parse(self):
if self.transform and getattr(self.transform, 'parse', None):
transform = self.transform.parse()
else:
transform = self.transform
return cde.BoundingBoxAugmentOperation(transform, self.ratio)
[docs]class CenterCrop(ImageTensorOperation, PyTensorOperation):
"""
Crop the input image at the center to the given size. If input image size is smaller than output size,
input image will be padded with 0 before cropping.
Args:
size (Union[int, sequence]): The output size of the cropped image.
If size is an integer, a square crop of size (size, size) is returned.
If size is a sequence of length 2, it should be (height, width).
The size value(s) must be larger than 0.
Raises:
TypeError: If `size` is not of type integer or sequence.
ValueError: If `size` is less than or equal to 0.
RuntimeError: If given tensor shape is not <H, W> or <..., H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>>
>>> # crop image to a square
>>> transforms_list1 = [vision.CenterCrop(50)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list1, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(50, 50, 3) uint8
>>>
>>> # crop image to portrait style
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list2 = [vision.CenterCrop((60, 40))]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list2, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(60, 40, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.CenterCrop(1)(data)
>>> print(output.shape, output.dtype)
(1, 1, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_center_crop
def __init__(self, size):
super().__init__()
if isinstance(size, int):
size = (size, size)
self.size = size
self.random = False
def parse(self):
return cde.CenterCropOperation(self.size)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be center cropped.
Returns:
PIL Image, cropped image.
"""
return util.center_crop(img, self.size)
[docs]class ConvertColor(ImageTensorOperation):
"""
Change the color space of the image.
Args:
convert_mode (ConvertMode): The mode of image channel conversion.
- ConvertMode.COLOR_BGR2BGRA, Convert BGR image to BGRA image.
- ConvertMode.COLOR_RGB2RGBA, Convert RGB image to RGBA image.
- ConvertMode.COLOR_BGRA2BGR, Convert BGRA image to BGR image.
- ConvertMode.COLOR_RGBA2RGB, Convert RGBA image to RGB image.
- ConvertMode.COLOR_BGR2RGBA, Convert BGR image to RGBA image.
- ConvertMode.COLOR_RGB2BGRA, Convert RGB image to BGRA image.
- ConvertMode.COLOR_RGBA2BGR, Convert RGBA image to BGR image.
- ConvertMode.COLOR_BGRA2RGB, Convert BGRA image to RGB image.
- ConvertMode.COLOR_BGR2RGB, Convert BGR image to RGB image.
- ConvertMode.COLOR_RGB2BGR, Convert RGB image to BGR image.
- ConvertMode.COLOR_BGRA2RGBA, Convert BGRA image to RGBA image.
- ConvertMode.COLOR_RGBA2BGRA, Convert RGBA image to BGRA image.
- ConvertMode.COLOR_BGR2GRAY, Convert BGR image to GRAY image.
- ConvertMode.COLOR_RGB2GRAY, Convert RGB image to GRAY image.
- ConvertMode.COLOR_GRAY2BGR, Convert GRAY image to BGR image.
- ConvertMode.COLOR_GRAY2RGB, Convert GRAY image to RGB image.
- ConvertMode.COLOR_GRAY2BGRA, Convert GRAY image to BGRA image.
- ConvertMode.COLOR_GRAY2RGBA, Convert GRAY image to RGBA image.
- ConvertMode.COLOR_BGRA2GRAY, Convert BGRA image to GRAY image.
- ConvertMode.COLOR_RGBA2GRAY, Convert RGBA image to GRAY image.
Raises:
TypeError: If `convert_mode` is not of type :class:`mindspore.dataset.vision.ConvertMode` .
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>>
>>> # Convert RGB images to GRAY images
>>> convert_op = vision.ConvertColor(vision.ConvertMode.COLOR_RGB2GRAY)
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=convert_op, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100) uint8
>>> # Convert RGB images to BGR images
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> convert_op = vision.ConvertColor(vision.ConvertMode.COLOR_RGB2BGR)
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=convert_op, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.ConvertColor(vision.ConvertMode.COLOR_RGB2GRAY)(data)
>>> print(output.shape, output.dtype)
(2, 2) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_convert_color
def __init__(self, convert_mode):
super().__init__()
self.convert_mode = convert_mode
self.implementation = Implementation.C
def parse(self):
return cde.ConvertColorOperation(ConvertMode.to_c_type(self.convert_mode))
[docs]class Crop(ImageTensorOperation):
"""
Crop the input image at a specific location.
Supports Ascend hardware acceleration and can be enabled through the `.device("Ascend")` method.
Args:
coordinates(sequence): Coordinates of the upper left corner of the cropping image. Must be a sequence of two
values, in the form of (top, left).
size (Union[int, sequence]): The output size of the cropped image.
If size is an integer, a square crop of size (size, size) is returned.
If size is a sequence of length 2, it should be (height, width).
The size value(s) must be larger than 0.
Raises:
TypeError: If `coordinates` is not of type sequence.
TypeError: If `size` is not of type integer or sequence.
ValueError: If `coordinates` is less than 0.
ValueError: If `size` is less than or equal to 0.
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> crop_op = vision.Crop((0, 0), 32)
>>> transforms_list = [crop_op]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(32, 32, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.Crop((0, 0), 1)(data)
>>> print(output.shape, output.dtype)
(1, 1, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_crop
def __init__(self, coordinates, size):
super().__init__()
if isinstance(size, int):
size = (size, size)
self.coordinates = coordinates
self.size = size
self.implementation = Implementation.C
[docs] @check_device_target
def device(self, device_target="CPU"):
"""
Set the device for the current operator execution.
- When the device is Ascend, input/output shape should be limited from [4, 6] to [32768, 32768].
Args:
device_target (str, optional): The operator will be executed on this device. Currently supports
``CPU`` and ``Ascend`` . Default: ``CPU`` .
Raises:
TypeError: If `device_target` is not of type str.
ValueError: If `device_target` is not within the valid set of ['CPU', 'Ascend'].
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> crop_op = vision.Crop((0, 0), (100, 75)).device("Ascend")
>>> transforms_list = [crop_op]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 75, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.Crop((0, 0), 64).device("Ascend")(data)
>>> print(output.shape, output.dtype)
(64, 64, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
self.device_target = device_target
return self
def parse(self):
return cde.CropOperation(self.coordinates, self.size, self.device_target)
[docs]class CutMixBatch(ImageTensorOperation):
"""
Apply CutMix transformation on input batch of images and labels.
Note that you need to make labels into one-hot format and batched before calling this operation.
Args:
image_batch_format (ImageBatchFormat): The method of padding. Can be any of
[ImageBatchFormat.NHWC, ImageBatchFormat.NCHW].
alpha (float, optional): Hyperparameter of beta distribution, must be larger than 0. Default: ``1.0``.
prob (float, optional): The probability by which CutMix is applied to each image,
which must be in range: [0.0, 1.0]. Default: ``1.0``.
Raises:
TypeError: If `image_batch_format` is not of type :class:`mindspore.dataset.vision.ImageBatchFormat` .
TypeError: If `alpha` is not of type float.
TypeError: If `prob` is not of type float.
ValueError: If `alpha` is less than or equal 0.
ValueError: If `prob` is not in range [0.0, 1.0].
RuntimeError: If given tensor shape is not <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.transforms as transforms
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import ImageBatchFormat
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(28, 28, 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"])
>>> onehot_op = transforms.OneHot(num_classes=10)
>>> numpy_slices_dataset= numpy_slices_dataset.map(operations=onehot_op, input_columns=["label"])
>>> cutmix_batch_op = vision.CutMixBatch(ImageBatchFormat.NHWC, 1.0, 0.5)
>>> numpy_slices_dataset = numpy_slices_dataset.batch(5)
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=cutmix_batch_op,
... 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
(5, 28, 28, 3) uint8
(5, 3, 10) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, (3, 3, 10, 10)).astype(np.uint8)
>>> label = np.array([[0, 1], [1, 0], [1, 0]])
>>> output = vision.CutMixBatch(vision.ImageBatchFormat.NCHW, 1.0, 1.0)(data, label)
>>> print(output[0].shape, output[0].dtype)
(3, 3, 10, 10) uint8
>>> print(output[1].shape, output[1].dtype)
(3, 2) float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_cut_mix_batch_c
def __init__(self, image_batch_format, alpha=1.0, prob=1.0):
super().__init__()
self.image_batch_format = image_batch_format.value
self.alpha = alpha
self.prob = prob
self.implementation = Implementation.C
def parse(self):
return cde.CutMixBatchOperation(ImageBatchFormat.to_c_type(self.image_batch_format), self.alpha, self.prob)
[docs]class CutOut(ImageTensorOperation):
"""
Randomly cut (mask) out a given number of square patches from the input image array.
Args:
length (int): The side length of each square patch, must be larger than 0.
num_patches (int, optional): Number of patches to be cut out of an image, must be larger than 0. Default: ``1``.
is_hwc (bool, optional): Whether the input image is in HWC format.
``True`` - HWC format, ``False`` - CHW format. Default: ``True``.
Raises:
TypeError: If `length` is not of type integer.
TypeError: If `is_hwc` is not of type bool.
TypeError: If `num_patches` is not of type integer.
ValueError: If `length` is less than or equal 0.
ValueError: If `num_patches` is less than or equal 0.
RuntimeError: If given tensor shape is not <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.CutOut(80, num_patches=10)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.CutOut(20)(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_cutout_new
def __init__(self, length, num_patches=1, is_hwc=True):
super().__init__()
self.length = length
self.num_patches = num_patches
self.is_hwc = is_hwc
self.random = False
self.implementation = Implementation.C
def parse(self):
return cde.CutOutOperation(self.length, self.num_patches, self.is_hwc)
[docs]class Decode(ImageTensorOperation, PyTensorOperation):
"""
Decode the input image in RGB mode.
Supported image formats: JPEG, BMP, PNG, TIFF, GIF(need `to_pil=True` ), WEBP(need `to_pil=True` ).
Supports Ascend hardware acceleration and can be enabled through the `.device("Ascend")` method.
Args:
to_pil (bool, optional): Whether to decode the image to the PIL data type. If ``True``,
the image will be decoded to the PIL data type, otherwise it will be decoded to the
NumPy data type. Default: ``False``.
Raises:
RuntimeError: If given tensor is not a 1D sequence.
RuntimeError: If the input is not raw image bytes.
RuntimeError: If the input image is already decoded.
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import os
>>> import numpy as np
>>> from PIL import Image, ImageDraw
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> class MyDataset:
... def __init__(self):
... self.data = []
... img = Image.new("RGB", (300, 300), (255, 255, 255))
... draw = ImageDraw.Draw(img)
... draw.ellipse(((0, 0), (100, 100)), fill=(255, 0, 0), outline=(255, 0, 0), width=5)
... img.save("./1.jpg")
... data = np.fromfile("./1.jpg", np.uint8)
... self.data.append(data)
...
... def __getitem__(self, index):
... return self.data[0]
...
... def __len__(self):
... return 5
>>>
>>> my_dataset = MyDataset()
>>> generator_dataset = ds.GeneratorDataset(my_dataset, column_names="image")
>>> transforms_list = [vision.Decode(), vision.RandomHorizontalFlip()]
>>> generator_dataset = generator_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in generator_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(300, 300, 3) uint8
>>> os.remove("./1.jpg")
>>>
>>> # Use the transform in eager mode
>>> img = Image.new("RGB", (300, 300), (255, 255, 255))
>>> draw = ImageDraw.Draw(img)
>>> draw.polygon([(50, 50), (150, 50), (100, 150)], fill=(0, 255, 0), outline=(0, 255, 0))
>>> img.save("./2.jpg")
>>> data = np.fromfile("./2.jpg", np.uint8)
>>> output = vision.Decode()(data)
>>> print(output.shape, output.dtype)
(300, 300, 3) uint8
>>> os.remove("./2.jpg")
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_decode
def __init__(self, to_pil=False):
super().__init__()
self.to_pil = to_pil
if to_pil:
self.implementation = Implementation.PY
else:
self.implementation = Implementation.C
def __call__(self, img):
"""
Call method for input conversion for eager mode with C++ implementation.
"""
if isinstance(img, bytes):
img = np.frombuffer(img, dtype=np.uint8)
if not isinstance(img, np.ndarray):
raise TypeError("The type of the encoded image should be {0}, but got {1}.".format(np.ndarray, type(img)))
if img.dtype.type is np.str_:
raise TypeError("The data type of the encoded image can not be {}.".format(img.dtype.type))
if img.ndim != 1:
raise TypeError("The number of array dimensions of the encoded image should be 1, "
"but got {0}.".format(img.ndim))
return super().__call__(img)
[docs] @check_device_target
def device(self, device_target="CPU"):
"""
Set the device for the current operator execution.
Args:
device_target (str, optional): The operator will be executed on this device. Currently supports
``CPU`` and ``Ascend`` . Default: ``CPU`` .
Raises:
TypeError: If `device_target` is not of type str.
ValueError: If `device_target` is not within the valid set of ['CPU', 'Ascend'].
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import os
>>> import numpy as np
>>> from PIL import Image, ImageDraw
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> class MyDataset:
... def __init__(self):
... self.data = []
... img = Image.new("RGB", (300, 300), (255, 255, 255))
... draw = ImageDraw.Draw(img)
... draw.ellipse(((0, 0), (100, 100)), fill=(255, 0, 0), outline=(255, 0, 0), width=5)
... img.save("./1.jpg")
... data = np.fromfile("./1.jpg", np.uint8)
... self.data.append(data)
...
... def __getitem__(self, index):
... return self.data[0]
...
... def __len__(self):
... return 5
>>>
>>> my_dataset = MyDataset()
>>> generator_dataset = ds.GeneratorDataset(my_dataset, column_names="image")
>>> decode_op = vision.Decode().device("Ascend")
>>> resize_op = vision.Resize([100, 75], Inter.BICUBIC)
>>> transforms_list = [decode_op, resize_op]
>>> generator_dataset = generator_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in generator_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 75, 3) uint8
>>> os.remove("./1.jpg")
>>>
>>> # Use the transform in eager mode
>>> img = Image.new("RGB", (300, 300), (255, 255, 255))
>>> draw = ImageDraw.Draw(img)
>>> draw.polygon([(50, 50), (150, 50), (100, 150)], fill=(0, 255, 0), outline=(0, 255, 0))
>>> img.save("./2.jpg")
>>> data = np.fromfile("./2.jpg", np.uint8)
>>> output = vision.Decode().device("Ascend")(data)
>>> print(output.shape, output.dtype)
(300, 300, 3) uint8
>>> os.remove("./2.jpg")
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
if self.implementation == Implementation.PY and device_target == "Ascend":
raise ValueError("The transform \"Decode(to_pil=True)\" cannot be performed on Ascend device, " +
"please set \"to_pil=False\".")
self.device_target = device_target
return self
def parse(self):
return cde.DecodeOperation(True, self.device_target)
def _execute_py(self, img):
"""
Execute method.
Args:
img (NumPy): Image to be decoded.
Returns:
img (NumPy, PIL Image), Decoded image.
"""
return util.decode(img)
[docs]class Equalize(ImageTensorOperation, PyTensorOperation):
"""
Apply histogram equalization on input image.
Raises:
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.Equalize()]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.Equalize()(data)
>>> print(output.shape, output.dtype)
(2, 2, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
def __init__(self):
super().__init__()
self.random = False
def parse(self):
return cde.EqualizeOperation()
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be equalized.
Returns:
PIL Image, equalized image.
"""
return util.equalize(img)
[docs]class Erase(ImageTensorOperation):
"""
Erase the input image with given value.
Args:
top (int): Vertical ordinate of the upper left corner of erased region.
left (int): Horizontal ordinate of the upper left corner of erased region.
height (int): Height of erased region.
width (int): Width of erased region.
value (Union[int, Sequence[int, int, int]], optional): Pixel value used to pad the erased area.
Default: ``0``. If int is provided, it will be used for all RGB channels.
If Sequence[int, int, int] is provided, it will be used for R, G, B channels respectively.
inplace (bool, optional): Whether to apply erasing inplace. Default: ``False``.
Raises:
TypeError: If `top` is not of type int.
ValueError: If `top` is negative.
TypeError: If `left` is not of type int.
ValueError: If `left` is negative.
TypeError: If `height` is not of type int.
ValueError: If `height` is not positive.
TypeError: If `width` is not of type int.
ValueError: If `width` is not positive.
TypeError: If `value` is not of type int or Sequence[int, int, int].
ValueError: If `value` is not in range of [0, 255].
TypeError: If `inplace` is not of type bool.
RuntimeError: If shape of the input image is not <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.Erase(10,10,10,10)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.Erase(0, 0, 2, 1)(data)
>>> print(output.shape, output.dtype)
(2, 2, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_erase
def __init__(self, top, left, height, width, value=0, inplace=False):
super().__init__()
self.top = top
self.left = left
self.height = height
self.width = width
if isinstance(value, int):
value = tuple([value] * 3)
self.value = value
self.inplace = inplace
def parse(self):
return cde.EraseOperation(self.top, self.left, self.height, self.width, self.value, self.inplace)
[docs]class FiveCrop(PyTensorOperation):
"""
Crop the given image into one central crop and four corners.
Args:
size (Union[int, Sequence[int, int]]): The size of the cropped image.
If a single integer is provided, a square of size (size, size) will be cropped with this value.
If a Sequence of length 2 is provided, an image of size (height, width) will be cropped.
Raises:
TypeError: If `size` is not of type integer or Sequence of integer.
ValueError: If `size` is not positive.
Supported Platforms:
``CPU``
Examples:
>>> import os
>>> import numpy as np
>>> from PIL import Image, ImageDraw
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.transforms import Compose
>>>
>>> # Use the transform in dataset pipeline mode
>>> class MyDataset:
... def __init__(self):
... self.data = []
... img = Image.new("RGB", (300, 300), (255, 255, 255))
... draw = ImageDraw.Draw(img)
... draw.ellipse(((0, 0), (100, 100)), fill=(255, 0, 0), outline=(255, 0, 0), width=5)
... img.save("./1.jpg")
... data = np.fromfile("./1.jpg", np.uint8)
... self.data.append(data)
...
... def __getitem__(self, index):
... return self.data[0]
...
... def __len__(self):
... return 5
>>>
>>> my_dataset = MyDataset()
>>> generator_dataset = ds.GeneratorDataset(my_dataset, column_names="image")
>>> transforms_list = Compose([vision.Decode(to_pil=True),
... vision.FiveCrop(size=200),
... # 4D stack of 5 images
... lambda *images: np.stack([vision.ToTensor()(image) for image in images])])
>>> # apply the transform to dataset through map function
>>> generator_dataset = generator_dataset.map(operations=transforms_list, input_columns="image")
>>> for item in generator_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(5, 3, 200, 200) float32
>>> os.remove("./1.jpg")
>>>
>>> # Use the transform in eager mode
>>> img = Image.new("RGB", (300, 300), (255, 255, 255))
>>> draw = ImageDraw.Draw(img)
>>> draw.polygon([(50, 50), (150, 50), (100, 150)], fill=(0, 255, 0), outline=(0, 255, 0))
>>> img.save("./2.jpg")
>>> data = Image.open("./2.jpg")
>>> output = vision.FiveCrop(size=20)(data)
>>> print(np.array(output).shape, np.array(output).dtype)
(5,) object
>>> os.remove("./2.jpg")
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_five_crop
def __init__(self, size):
super().__init__()
self.size = size
self.random = False
self.implementation = Implementation.PY
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be cropped.
Returns:
tuple, a tuple of five PIL Image in order of top_left, top_right, bottom_left, bottom_right, center.
"""
return util.five_crop(img, self.size)
[docs]class GaussianBlur(ImageTensorOperation):
r"""
Blur input image with the specified Gaussian kernel.
Supports Ascend hardware acceleration and can be enabled through the `.device("Ascend")` method.
Args:
kernel_size (Union[int, Sequence[int, int]]): The size of the Gaussian kernel. Must be positive and odd.
If the input type is int, the value will be used as both the width and height of the Gaussian kernel.
If the input type is Sequence[int, int], the two elements will be used as the width and height of the
Gaussian kernel respectively.
sigma (Union[float, Sequence[float, float]], optional): The standard deviation of the Gaussian kernel.
Must be positive.
If the input type is float, the value will be used as the standard deviation of both the width and
height of the Gaussian kernel.
If the input type is Sequence[float, float], the two elements will be used as the standard deviation
of the width and height of the Gaussian kernel respectively.
Default: ``None`` , the standard deviation of the Gaussian kernel will be obtained by the
formula :math:`((kernel\_size - 1) * 0.5 - 1) * 0.3 + 0.8` .
Raises:
TypeError: If `kernel_size` is not of type int or Sequence[int].
TypeError: If `sigma` is not of type float or Sequence[float].
ValueError: If `kernel_size` is not positive and odd.
ValueError: If `sigma` is not positive.
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.GaussianBlur(3, 3)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.GaussianBlur(3, 3)(data)
>>> print(output.shape, output.dtype)
(2, 2, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_gaussian_blur
def __init__(self, kernel_size, sigma=None):
super().__init__()
if isinstance(kernel_size, int):
kernel_size = (kernel_size,)
if sigma is None:
sigma = (0,)
elif isinstance(sigma, (int, float)):
sigma = (float(sigma),)
self.kernel_size = kernel_size
self.sigma = sigma
self.implementation = Implementation.C
[docs] @check_device_target
def device(self, device_target="CPU"):
"""
Set the device for the current operator execution.
- When the device is Ascend, the parameter `kernel_size` only supports values 1, 3, and 5.
input shape should be limited from [4, 6] to [8192, 4096].
Args:
device_target (str, optional): The operator will be executed on this device. Currently supports
``CPU`` and ``Ascend`` . Default: ``CPU`` .
Raises:
TypeError: If `device_target` is not of type str.
ValueError: If `device_target` is not within the valid set of ['CPU', 'Ascend'].
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> blur_op = vision.GaussianBlur(3, 3).device("Ascend")
>>> transforms_list = [blur_op]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.GaussianBlur(3, 3).device("Ascend")(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
self.device_target = device_target
if device_target == "Ascend":
for k in self.kernel_size:
if k not in [1, 3, 5]:
raise RuntimeError("When target is Ascend, `kernel_size` only supports values 1, 3, and 5.")
return self
def parse(self):
return cde.GaussianBlurOperation(self.kernel_size, self.sigma, self.device_target)
[docs]class Grayscale(PyTensorOperation):
"""
Convert the input PIL Image to grayscale.
Args:
num_output_channels (int): The number of channels desired for the output image, must be ``1`` or ``3``.
If ``3`` is provided, the returned image will have 3 identical RGB channels. Default: ``1``.
Raises:
TypeError: If `num_output_channels` is not of type integer.
ValueError: If `num_output_channels` is not ``1`` or ``3``.
Supported Platforms:
``CPU``
Examples:
>>> import os
>>> import numpy as np
>>> from PIL import Image, ImageDraw
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.transforms import Compose
>>>
>>> # Use the transform in dataset pipeline mode
>>> class MyDataset:
... def __init__(self):
... self.data = []
... img = Image.new("RGB", (300, 300), (255, 255, 255))
... draw = ImageDraw.Draw(img)
... draw.ellipse(((0, 0), (100, 100)), fill=(255, 0, 0), outline=(255, 0, 0), width=5)
... img.save("./1.jpg")
... data = np.fromfile("./1.jpg", np.uint8)
... self.data.append(data)
...
... def __getitem__(self, index):
... return self.data[0]
...
... def __len__(self):
... return 5
>>>
>>> my_dataset = MyDataset()
>>> generator_dataset = ds.GeneratorDataset(my_dataset, column_names="image")
>>> transforms_list = Compose([vision.Decode(to_pil=True),
... vision.Grayscale(3),
... vision.ToTensor()])
>>> # apply the transform to dataset through map function
>>> generator_dataset = generator_dataset.map(operations=transforms_list, input_columns="image")
>>> for item in generator_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(3, 300, 300) float32
>>> os.remove("./1.jpg")
>>>
>>> # Use the transform in eager mode
>>> img = Image.new("RGB", (300, 300), (255, 255, 255))
>>> draw = ImageDraw.Draw(img)
>>> draw.polygon([(50, 50), (150, 50), (100, 150)], fill=(0, 255, 0), outline=(0, 255, 0))
>>> img.save("./2.jpg")
>>> data = Image.open("./2.jpg")
>>> output = vision.Grayscale(3)(data)
>>> print(np.array(output).shape, np.array(output).dtype)
(300, 300, 3) uint8
>>> os.remove("./2.jpg")
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_num_channels
def __init__(self, num_output_channels=1):
super().__init__()
self.num_output_channels = num_output_channels
self.random = False
self.implementation = Implementation.PY
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be converted to grayscale.
Returns:
PIL Image, converted grayscale image.
"""
return util.grayscale(img, num_output_channels=self.num_output_channels)
[docs]class HorizontalFlip(ImageTensorOperation):
"""
Flip the input image horizontally.
Raises:
RuntimeError: If given tensor shape is not <H, W> or <..., H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.HorizontalFlip()]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.HorizontalFlip()(data)
>>> print(output.shape, output.dtype)
(2, 2, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
def __init__(self):
super().__init__()
self.implementation = Implementation.C
[docs] @check_device_target
def device(self, device_target="CPU"):
"""
Set the device for the current operator execution.
- When the device is Ascend, input shape should be limited from [4, 6] to [8192, 4096].
Args:
device_target (str, optional): The operator will be executed on this device. Currently supports
``CPU`` and ``Ascend`` . Default: ``CPU`` .
Raises:
TypeError: If `device_target` is not of type str.
ValueError: If `device_target` is not within the valid set of ['CPU', 'Ascend'].
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> horizontal_flip_op = vision.HorizontalFlip().device("Ascend")
>>> transforms_list = [horizontal_flip_op]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.HorizontalFlip().device("Ascend")(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
self.device_target = device_target
return self
def parse(self):
return cde.HorizontalFlipOperation(self.device_target)
[docs]class HsvToRgb(PyTensorOperation):
"""
Convert the input numpy.ndarray images from HSV to RGB.
Args:
is_hwc (bool): If ``True``, means the input image is in shape of <H, W, C> or <N, H, W, C>.
Otherwise, it is in shape of <C, H, W> or <N, C, H, W>. Default: ``False``.
Raises:
TypeError: If `is_hwc` is not of type bool.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.transforms import Compose
>>>
>>> # Use the transform in dataset pipeline mode
>>> transforms_list = Compose([vision.CenterCrop(20),
... vision.ToTensor(),
... vision.HsvToRgb()])
>>> # apply the transform to dataset through map function
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns="image")
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(3, 20, 20) float64
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.HsvToRgb(is_hwc=True)(data)
>>> print(output.shape, output.dtype)
(2, 2, 3) float64
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_hsv_to_rgb
def __init__(self, is_hwc=False):
super().__init__()
self.is_hwc = is_hwc
self.random = False
self.implementation = Implementation.PY
def _execute_py(self, hsv_imgs):
"""
Execute method.
Args:
hsv_imgs (numpy.ndarray): HSV images to be converted.
Returns:
numpy.ndarray, converted RGB images.
"""
return util.hsv_to_rgbs(hsv_imgs, self.is_hwc)
[docs]class HWC2CHW(ImageTensorOperation):
"""
Transpose the input image from shape <H, W, C> to <C, H, W>.
If the input image is of shape <H, W>, it will remain unchanged.
Note:
This operation is executed on the CPU by default, but it is also supported
to be executed on the GPU or Ascend via heterogeneous acceleration.
Raises:
RuntimeError: If shape of the input image is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU`` ``GPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.RandomHorizontalFlip(0.75),
... vision.RandomCrop(64),
... vision.HWC2CHW()]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(3, 64, 64) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.HWC2CHW()(data)
>>> print(output.shape, output.dtype)
(3, 2, 2) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
def __init__(self):
super().__init__()
self.implementation = Implementation.C
self.random = False
def parse(self):
return cde.HwcToChwOperation()
[docs]class Invert(ImageTensorOperation, PyTensorOperation):
"""
Invert the colors of the input RGB image.
For each pixel in the image, if the original pixel value is `pixel`,
the inverted pixel value will be `255 - pixel`.
Raises:
RuntimeError: If the input image is not in shape of <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.Invert()]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 2, 3))
>>> output = vision.Invert()(data)
>>> print(output.shape, output.dtype)
(2, 2, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
def __init__(self):
super().__init__()
self.random = False
def parse(self):
return cde.InvertOperation()
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be color inverted.
Returns:
PIL Image, color inverted image.
"""
return util.invert_color(img)
[docs]class MixUp(PyTensorOperation):
"""
Randomly mix up a batch of numpy.ndarray images together with its labels.
Each image will be multiplied by a random weight :math:`lambda` generated from the Beta distribution and then added
to another image multiplied by :math:`1 - lambda`. The same transformation will be applied to their labels with the
same value of :math:`lambda`. Make sure that the labels are one-hot encoded in advance.
Args:
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:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_mix_up
def __init__(self, batch_size, alpha, is_single=True):
super().__init__()
self.image = 0
self.label = 0
self.is_first = True
self.batch_size = batch_size
self.alpha = alpha
self.is_single = is_single
self.random = False
self.implementation = Implementation.PY
def __call__(self, image, label):
"""
Call method to apply mix up transformation to image and label.
Note: No execute method for MixUp
Args:
image (numpy.ndarray): Images to be mixed up.
label (numpy.ndarray): Labels to be mixed up.
Returns:
numpy.ndarray, images after mixing up.
numpy.ndarray, labels after mixing up.
"""
if self.is_single:
return util.mix_up_single(self.batch_size, image, label, self.alpha)
return util.mix_up_muti(self, self.batch_size, image, label, self.alpha)
[docs]class MixUpBatch(ImageTensorOperation):
"""
Apply MixUp transformation on input batch of images and labels. Each image is
multiplied by a random weight (lambda) and then added to a randomly selected image from the batch
multiplied by (1 - lambda). The same formula is also applied to the one-hot labels.
The lambda is generated based on the specified alpha value. Two coefficients x1, x2 are randomly generated
in the range [alpha, 1], and lambda = (x1 / (x1 + x2)).
Note that you need to make labels into one-hot format and batched before calling this operation.
Args:
alpha (float, optional): Hyperparameter of beta distribution. The value must be positive.
Default: ``1.0``.
Raises:
TypeError: If `alpha` is not of type float.
ValueError: If `alpha` is not positive.
RuntimeError: If given tensor shape is not <N, H, W, C> or <N, C, H, W>.
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"])
>>> onehot_op = transforms.OneHot(num_classes=10)
>>> numpy_slices_dataset= numpy_slices_dataset.map(operations=onehot_op,
... input_columns=["label"])
>>> mixup_batch_op = vision.MixUpBatch(alpha=0.9)
>>> numpy_slices_dataset = numpy_slices_dataset.batch(5)
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=mixup_batch_op,
... 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
(5, 64, 64, 3) uint8
(5, 3, 10) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, (2, 10, 10, 3)).astype(np.uint8)
>>> label = np.array([[0, 1], [1, 0]])
>>> output = vision.MixUpBatch(1)(data, label)
>>> print(output[0].shape, output[0].dtype)
(2, 10, 10, 3) uint8
>>> print(output[1].shape, output[1].dtype)
(2, 2) float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_mix_up_batch_c
def __init__(self, alpha=1.0):
super().__init__()
self.alpha = alpha
self.implementation = Implementation.C
def parse(self):
return cde.MixUpBatchOperation(self.alpha)
[docs]class Normalize(ImageTensorOperation):
"""
Normalize the input image with respect to mean and standard deviation. This operation will normalize
the input image with: output[channel] = (input[channel] - mean[channel]) / std[channel], where channel >= 1.
Supports Ascend hardware acceleration and can be enabled through the `.device("Ascend")` method.
Note:
This operation is executed on the CPU by default, but it is also supported
to be executed on the GPU or Ascend via heterogeneous acceleration.
Args:
mean (sequence): List or tuple of mean values for each channel, with respect to channel order.
The mean values must be in range [0.0, 255.0].
std (sequence): List or tuple of standard deviations for each channel, with respect to channel order.
The standard deviation values must be in range (0.0, 255.0].
is_hwc (bool, optional): Whether the input image is HWC.
``True`` - HWC format, ``False`` - CHW format. Default: ``True``.
Raises:
TypeError: If `mean` is not of type sequence.
TypeError: If `std` is not of type sequence.
TypeError: If `is_hwc` is not of type bool.
ValueError: If `mean` is not in range [0.0, 255.0].
ValueError: If `std` is not in range (0.0, 255.0].
RuntimeError: If given tensor format is not <H, W> or <..., H, W, C>.
Supported Platforms:
``CPU`` ``GPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> normalize_op = vision.Normalize(mean=[121.0, 115.0, 100.0], std=[70.0, 68.0, 71.0], is_hwc=True)
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=[normalize_op],
... input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.Normalize(mean=[121.0, 115.0, 100.0], std=[70.0, 68.0, 71.0])(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_normalize
def __init__(self, mean, std, is_hwc=True):
super().__init__()
self.mean = mean
self.std = std
self.is_hwc = is_hwc
self.random = False
self.implementation = Implementation.C
[docs] @check_device_target
def device(self, device_target="CPU"):
"""
Set the device for the current operator execution.
- When the device is CPU, input type support `uint8`/`float32`/`float64`, input channel support 1/2/3.
- When the device is Ascend, input type supports `uint8`/`float32`, input channel supports 1/3.
input shape should be limited from [4, 6] to [8192, 4096].
Args:
device_target (str, optional): The operator will be executed on this device. Currently supports
``CPU`` and ``Ascend`` . Default: ``CPU`` .
Raises:
TypeError: If `device_target` is not of type str.
ValueError: If `device_target` is not within the valid set of ['CPU', 'Ascend'].
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> resize_op = vision.Resize([100, 75], Inter.BICUBIC)
>>> transforms_list = [resize_op]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> normalize_op = vision.Normalize(mean=[121.0, 115.0, 100.0], std=[70.0, 68.0, 71.0]).device("Ascend")
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=normalize_op, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 75, 3) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.Normalize(mean=[121.0, 115.0, 100.0], std=[70.0, 68.0, 71.0]).device("Ascend")(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
self.device_target = device_target
return self
def parse(self):
return cde.NormalizeOperation(self.mean, self.std, self.is_hwc, self.device_target)
[docs]class NormalizePad(ImageTensorOperation):
"""
Normalize the input image with respect to mean and standard deviation then pad an extra channel with value zero.
Args:
mean (sequence): List or tuple of mean values for each channel, with respect to channel order.
The mean values must be in range (0.0, 255.0].
std (sequence): List or tuple of standard deviations for each channel, with respect to channel order.
The standard deviation values must be in range (0.0, 255.0].
dtype (str, optional): Set the output data type of normalized image. Default: ``"float32"``.
is_hwc (bool, optional): Specify the format of input image.
``True`` - HW(C) format, ``False`` - CHW format. Default: ``True``.
Raises:
TypeError: If `mean` is not of type sequence.
TypeError: If `std` is not of type sequence.
TypeError: If `dtype` is not of type string.
TypeError: If `is_hwc` is not of type bool.
ValueError: If `mean` is not in range [0.0, 255.0].
ValueError: If `mean` is not in range (0.0, 255.0].
RuntimeError: If given tensor shape is not <H, W>, <H, W, C> or <C, H, W>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> normalize_pad_op = vision.NormalizePad(mean=[121.0, 115.0, 100.0],
... std=[70.0, 68.0, 71.0],
... dtype="float32")
>>> transforms_list = [normalize_pad_op]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 4) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.NormalizePad(mean=[121.0, 115.0, 100.0], std=[70.0, 68.0, 71.0], dtype="float32")(data)
>>> print(output.shape, output.dtype)
(100, 100, 4) float32
"""
@check_normalizepad
def __init__(self, mean, std, dtype="float32", is_hwc=True):
super().__init__()
self.mean = mean
self.std = std
self.dtype = dtype
self.is_hwc = is_hwc
self.random = False
self.implementation = Implementation.C
def parse(self):
return cde.NormalizePadOperation(self.mean, self.std, self.dtype, self.is_hwc)
[docs]class Pad(ImageTensorOperation, PyTensorOperation):
"""
Pad the image according to padding parameters.
Supports Ascend hardware acceleration and can be enabled through the `.device("Ascend")` method.
Args:
padding (Union[int, Sequence[int, int], Sequence[int, int, int, int]]): The number of pixels
to pad each border of the image.
If a single number is provided, it pads all borders with this value.
If a tuple or lists of 2 values are provided, it pads the (left and right)
with the first value and (top and bottom) with the second value.
If 4 values are provided as a list or tuple, it pads the left, top, right and bottom respectively.
The pad values must be non-negative.
fill_value (Union[int, tuple[int]], optional): The pixel intensity of the borders, only valid for
`padding_mode` ``Border.CONSTANT``. 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``.
padding_mode (Border, optional): The method of padding. Default: ``Border.CONSTANT``. Can be
``Border.CONSTANT``, ``Border.EDGE``, ``Border.REFLECT``, ``Border.SYMMETRIC``.
- ``Border.CONSTANT`` , means it fills the border with constant values.
- ``Border.EDGE`` , means it pads with the last value on the edge.
- ``Border.REFLECT`` , means it reflects the values on the edge omitting the last
value of edge.
- ``Border.SYMMETRIC`` , means it reflects the values on the edge repeating the last
value of edge.
Raises:
TypeError: If `padding` is not of type int or Sequence[int, int], Sequence[int, int, int, int].
TypeError: If `fill_value` is not of type int or tuple[int].
TypeError: If `padding_mode` is not of type :class:`mindspore.dataset.vision.Border` .
ValueError: If `padding` is negative.
ValueError: If `fill_value` is not in range [0, 255].
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.Pad([100, 100, 100, 100])]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(300, 300, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.Pad([100, 100, 100, 100])(data)
>>> print(output.shape, output.dtype)
(300, 300, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_pad
def __init__(self, padding, fill_value=0, padding_mode=Border.CONSTANT):
super().__init__()
padding = parse_padding(padding)
if isinstance(fill_value, int):
fill_value = tuple([fill_value] * 3)
self.padding = padding
self.fill_value = fill_value
self.random = False
self.padding_mode = padding_mode
[docs] @check_device_target
def device(self, device_target="CPU"):
"""
Set the device for the current operator execution.
- When the device is Ascend, input/output shape should be limited from [4, 6] to [32768, 32768].
Args:
device_target (str, optional): The operator will be executed on this device. Currently supports
``CPU`` and ``Ascend`` . Default: ``CPU`` .
Raises:
TypeError: If `device_target` is not of type str.
ValueError: If `device_target` is not within the valid set of ['CPU', 'Ascend'].
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> pad_op = vision.Pad([100, 100, 100, 100]).device("Ascend")
>>> transforms_list = [pad_op]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(300, 300, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.Pad([100, 100, 100, 100]).device("Ascend")(data)
>>> print(output.shape, output.dtype)
(300, 300, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
self.device_target = device_target
return self
def parse(self):
return cde.PadOperation(self.padding, self.fill_value, Border.to_c_type(self.padding_mode), self.device_target)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be padded.
Returns:
PIL Image, padded image.
"""
return util.pad(img, self.padding, self.fill_value, Border.to_python_type(self.padding_mode))
[docs]class PadToSize(ImageTensorOperation):
"""
Pad the image to a fixed size.
Args:
size (Union[int, Sequence[int, int]]): The target size to pad.
If int is provided, pad the image to [size, size].
If Sequence[int, int] is provided, it should be in order of [height, width].
offset (Union[int, Sequence[int, int]], optional): The lengths to pad on the top and left.
If int is provided, pad both top and left borders with this value.
If Sequence[int, int] is provided, is should be in order of [top, left].
Default: ``None``, means to pad symmetrically, keeping the original image in center.
fill_value (Union[int, tuple[int, int, int]], optional): Pixel value used to pad the borders,
only valid when `padding_mode` is ``Border.CONSTANT``.
If int is provided, it will be used for all RGB channels.
If tuple[int, int, int] is provided, it will be used for R, G, B channels respectively. Default: 0.
padding_mode (Border, optional): Method of padding. It can be ``Border.CONSTANT``, ``Border.EDGE``,
``Border.REFLECT`` or Border.SYMMETRIC. Default: ``Border.CONSTANT``.
- ``Border.CONSTANT`` , pads with a constant value.
- ``Border.EDGE`` , pads with the last value at the edge of the image.
- ``Border.REFLECT`` , pads with reflection of the image omitting the last value on the edge.
- ``Border.SYMMETRIC`` , pads with reflection of the image repeating the last value on the edge.
Raises:
TypeError: If `size` is not of type int or Sequence[int, int].
TypeError: If `offset` is not of type int or Sequence[int, int].
TypeError: If `fill_value` is not of type int or tuple[int, int, int].
TypeError: If `padding_mode` is not of type :class:`mindspore.dataset.vision.Border` .
ValueError: If `size` is not positive.
ValueError: If `offset` is negative.
ValueError: If `fill_value` is not in range of [0, 255].
RuntimeError: If shape of the input image is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.PadToSize([256, 256])]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(256, 256, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.PadToSize([256, 256])(data)
>>> print(output.shape, output.dtype)
(256, 256, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_pad_to_size
def __init__(self, size, offset=None, fill_value=0, padding_mode=Border.CONSTANT):
super().__init__()
self.size = [size, size] if isinstance(size, int) else size
if offset is None:
self.offset = []
else:
self.offset = [offset, offset] if isinstance(offset, int) else offset
self.fill_value = tuple([fill_value] * 3) if isinstance(fill_value, int) else fill_value
self.padding_mode = padding_mode
self.implementation = Implementation.C
def parse(self):
return cde.PadToSizeOperation(self.size, self.offset, self.fill_value, Border.to_c_type(self.padding_mode))
[docs]class Perspective(ImageTensorOperation, PyTensorOperation):
"""
Apply perspective transformation on input image.
Args:
start_points (Sequence[Sequence[int, int]]): Sequence of the starting point coordinates, containing four
two-element subsequences, corresponding to [top-left, top-right, bottom-right, bottom-left] of the
quadrilateral in the original image.
end_points (Sequence[Sequence[int, int]]): Sequence of the ending point coordinates, containing four
two-element subsequences, corresponding to [top-left, top-right, bottom-right, bottom-left] of the
quadrilateral in the target image.
interpolation (Inter, optional): Image interpolation method defined by :class:`~.vision.Inter` .
Default: ``Inter.BILINEAR``.
Raises:
TypeError: If `start_points` is not of type Sequence[Sequence[int, int]].
TypeError: If `end_points` is not of type Sequence[Sequence[int, int]].
TypeError: If `interpolation` is not of type :class:`~.vision.Inter` .
RuntimeError: If shape of the input image is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> start_points = [[0, 63], [63, 63], [63, 0], [0, 0]]
>>> end_points = [[0, 32], [32, 32], [32, 0], [0, 0]]
>>> transforms_list = [vision.Perspective(start_points, end_points, Inter.BILINEAR)]
>>> # apply the transform to dataset through map function
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns="image")
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> start_points = [[0, 63], [63, 63], [63, 0], [0, 0]]
>>> end_points = [[0, 32], [32, 32], [32, 0], [0, 0]]
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.Perspective(start_points, end_points, Inter.BILINEAR)(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_perspective
def __init__(self, start_points, end_points, interpolation=Inter.BILINEAR):
super().__init__()
self.start_points = start_points
self.end_points = end_points
self.interpolation = interpolation
if interpolation in [Inter.AREA, Inter.PILCUBIC]:
self.implementation = Implementation.C
elif interpolation == Inter.ANTIALIAS:
self.implementation = Implementation.PY
self.random = False
[docs] @check_device_target
def device(self, device_target="CPU"):
"""
Set the device for the current operator execution.
- When the device is Ascend, input shape should be limited from [6, 10] to [8192, 4096].
Args:
device_target (str, optional): The operator will be executed on this device. Currently supports
``CPU`` and ``Ascend`` . Default: ``CPU`` .
Raises:
TypeError: If `device_target` is not of type str.
ValueError: If `device_target` is not within the valid set of ['CPU', 'Ascend'].
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> start_points = [[0, 63], [63, 63], [63, 0], [0, 0]]
>>> end_points = [[0, 32], [32, 32], [32, 0], [0, 0]]
>>> perspective_op = vision.Perspective(start_points, end_points).device("Ascend")
>>> transforms_list = [perspective_op]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> start_points = [[0, 63], [63, 63], [63, 0], [0, 0]]
>>> end_points = [[0, 32], [32, 32], [32, 0], [0, 0]]
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.Perspective(start_points, end_points, Inter.BILINEAR).device("Ascend")(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
self.device_target = device_target
return self
def parse(self):
if self.interpolation == Inter.ANTIALIAS:
raise TypeError("Current Interpolation is not supported with NumPy input.")
return cde.PerspectiveOperation(self.start_points, self.end_points,
Inter.to_c_type(self.interpolation), self.device_target)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be perspectived.
Returns:
PIL Image, perspectived image.
"""
if self.interpolation in [Inter.AREA, Inter.PILCUBIC]:
raise TypeError("Current Interpolation is not supported with PIL input.")
return util.perspective(img, self.start_points, self.end_points, Inter.to_python_type(self.interpolation))
[docs]class Posterize(ImageTensorOperation):
"""
Reduce the bit depth of the color channels of image to create a high contrast and vivid color effect,
similar to that seen in posters or printed materials.
Args:
bits (int): The number of bits to keep for each channel, should be in range of [0, 8].
Raises:
TypeError: If `bits` is not of type int.
ValueError: If `bits` is not in range [0, 8].
RuntimeError: If shape of the input image is not <H, W> or <H, W, C>.
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.Posterize(4)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.Posterize(4)(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_posterize
def __init__(self, bits):
super().__init__()
self.bits = bits
self.implementation = Implementation.C
def parse(self):
return cde.PosterizeOperation(self.bits)
[docs]class RandAugment(ImageTensorOperation):
"""
Apply RandAugment data augmentation method on the input image.
Refer to `RandAugment: Learning Augmentation Strategies from Data <https://arxiv.org/pdf/1909.13719.pdf>`_ .
Only support 3-channel RGB image.
Args:
num_ops (int, optional): Number of augmentation transformations to apply sequentially. Default: ``2``.
magnitude (int, optional): Magnitude for all the transformations, must be smaller than
`num_magnitude_bins`. Default: ``9``.
num_magnitude_bins (int, optional): The number of different magnitude values,
must be no less than 2. Default: ``31``.
interpolation (Inter, optional): Image interpolation method defined by :class:`~.vision.Inter` .
Default: ``Inter.NEAREST``.
fill_value (Union[int, tuple[int, int, int]], optional): Pixel fill value for the area outside the
transformed image, must be in range of [0, 255]. Default: ``0``.
If int is provided, pad all RGB channels with this value.
If tuple[int, int, int] is provided, pad R, G, B channels respectively.
Raises:
TypeError: If `num_ops` is not of type int.
ValueError: If `num_ops` is negative.
TypeError: If `magnitude` is not of type int.
ValueError: If `magnitude` is not positive.
TypeError: If `num_magnitude_bins` is not of type int.
ValueError: If `num_magnitude_bins` is less than 2.
TypeError: If `interpolation` not of type :class:`~.vision.Inter` .
TypeError: If `fill_value` is not of type int or tuple[int, int, int].
ValueError: If `fill_value` is not in range of [0, 255].
RuntimeError: If shape of the input image is not <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.RandAugment()]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandAugment(interpolation=Inter.BILINEAR, fill_value=255)(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_rand_augment
def __init__(self, num_ops=2, magnitude=9, num_magnitude_bins=31, interpolation=Inter.NEAREST, fill_value=0):
super().__init__()
self.num_ops = num_ops
self.magnitude = magnitude
self.num_magnitude_bins = num_magnitude_bins
self.interpolation = interpolation
if isinstance(fill_value, int):
fill_value = tuple([fill_value] * 3)
self.fill_value = fill_value
self.implementation = Implementation.C
def parse(self):
return cde.RandAugmentOperation(self.num_ops, self.magnitude, self.num_magnitude_bins,
Inter.to_c_type(self.interpolation), self.fill_value)
[docs]class RandomAdjustSharpness(ImageTensorOperation):
"""
Randomly adjust the sharpness of the input image with a given probability.
Args:
degree (float): Sharpness adjustment degree, which must be non negative.
Degree of ``0.0`` gives a blurred image, degree of ``1.0`` gives the original image,
and degree of ``2.0`` increases the sharpness by a factor of 2.
prob (float, optional): Probability of the image being sharpness adjusted, which
must be in range of [0.0, 1.0]. Default: ``0.5``.
Raises:
TypeError: If `degree` is not of type float.
TypeError: If `prob` is not of type float.
ValueError: If `degree` is negative.
ValueError: If `prob` is not in range [0.0, 1.0].
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.RandomAdjustSharpness(2.0, 0.5)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomAdjustSharpness(2.0, 1.0)(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_random_adjust_sharpness
def __init__(self, degree, prob=0.5):
super().__init__()
self.prob = prob
self.degree = degree
self.implementation = Implementation.C
def parse(self):
return cde.RandomAdjustSharpnessOperation(self.degree, self.prob)
[docs]class RandomAffine(ImageTensorOperation, PyTensorOperation):
"""
Apply Random affine transformation to the input image.
Args:
degrees (Union[int, float, sequence]): Range of the rotation degrees.
If `degrees` is a number, the range will be (-degrees, degrees).
If `degrees` is a sequence, it should be (min, max).
translate (sequence, optional): Sequence (tx_min, tx_max, ty_min, ty_max) of minimum/maximum translation in
x(horizontal) and y(vertical) directions, range [-1.0, 1.0]. Default: ``None``.
The horizontal and vertical shift is selected randomly from the range:
(tx_min*width, tx_max*width) and (ty_min*height, ty_max*height), respectively.
If a tuple or list of size 2, then a translate parallel to the X axis in the range of
(translate[0], translate[1]) is applied.
If a tuple or list of size 4, then a translate parallel to the X axis in the range of
(translate[0], translate[1]) and a translate parallel to the Y axis in the range of
(translate[2], translate[3]) are applied.
If ``None``, no translation is applied.
scale (sequence, optional): Scaling factor interval, which must be non negative.
Default: ``None``, original scale is used.
shear (Union[float, Sequence[float, float], Sequence[float, float, float, float]], optional):
Range of shear factor to select from.
If float is provided, a shearing parallel to X axis with a factor selected from
( `-shear` , `shear` ) will be applied.
If Sequence[float, float] is provided, a shearing parallel to X axis with a factor selected
from ( `shear` [0], `shear` [1]) will be applied.
If Sequence[float, float, float, float] is provided, a shearing parallel to X axis with a factor selected
from ( `shear` [0], `shear` [1]) and a shearing parallel to Y axis with a factor selected from
( `shear` [2], `shear` [3]) will be applied. Default: ``None``, means no shearing.
resample (Inter, optional): Image interpolation method defined by :class:`~.vision.Inter` .
Default: ``Inter.NEAREST``.
fill_value (Union[int, tuple[int]], optional): Optional fill_value to fill the area outside the transform
in the output image. There must be three elements in tuple and the value of single element is [0, 255].
Default: ``0``, filling is performed.
Raises:
TypeError: If `degrees` is not of type int, float or sequence.
TypeError: If `translate` is not of type sequence.
TypeError: If `scale` is not of type sequence.
TypeError: If `shear` is not of type int, float or sequence.
TypeError: If `resample` is not of type :class:`~.vision.Inter` .
TypeError: If `fill_value` is not of type int or tuple[int].
ValueError: If `degrees` is negative.
ValueError: If `translate` is not in range [-1.0, 1.0].
ValueError: If `scale` is negative.
ValueError: If `shear` is not positive.
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> random_affine_op = vision.RandomAffine(degrees=15,
... translate=(-0.1, 0.1, 0, 0),
... scale=(0.9, 1.1),
... resample=Inter.NEAREST)
>>> transforms_list = [random_affine_op]
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomAffine(degrees=15, translate=(-0.1, 0.1, 0, 0),
... scale=(0.9, 1.1), resample=Inter.NEAREST)(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_random_affine
def __init__(self, degrees, translate=None, scale=None, shear=None, resample=Inter.NEAREST, fill_value=0):
super().__init__()
# Parameter checking
if shear is not None:
if isinstance(shear, numbers.Number):
shear = (-1 * shear, shear, 0., 0.)
else:
if len(shear) == 2:
shear = [shear[0], shear[1], 0., 0.]
elif len(shear) == 4:
shear = [s for s in shear]
if isinstance(degrees, numbers.Number):
degrees = (-1 * degrees, degrees)
if isinstance(fill_value, numbers.Number):
fill_value = (fill_value, fill_value, fill_value)
# translation
if translate is None:
translate = (0.0, 0.0, 0.0, 0.0)
# scale
if scale is None:
scale = (1.0, 1.0)
# shear
if shear is None:
shear = (0.0, 0.0, 0.0, 0.0)
self.degrees = degrees
self.translate = translate
self.scale = scale
self.shear = shear
self.resample = resample
if resample in [Inter.AREA, Inter.PILCUBIC]:
self.implementation = Implementation.C
elif resample == Inter.ANTIALIAS:
self.implementation = Implementation.PY
self.fill_value = fill_value
def parse(self):
if self.resample == Inter.ANTIALIAS:
raise TypeError("Current Interpolation is not supported with NumPy input.")
return cde.RandomAffineOperation(self.degrees, self.translate, self.scale, self.shear,
Inter.to_c_type(self.resample), self.fill_value)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be randomly affine transformed.
Returns:
PIL Image, randomly affine transformed image.
"""
if self.resample in [Inter.AREA, Inter.PILCUBIC]:
raise TypeError("Current Interpolation is not supported with PIL input.")
return util.random_affine(img,
self.degrees,
self.translate,
self.scale,
self.shear,
Inter.to_python_type(self.resample),
self.fill_value)
[docs]class RandomAutoContrast(ImageTensorOperation):
"""
Automatically adjust the contrast of the image with a given probability.
Args:
cutoff (float, optional): Percent of the lightest and darkest pixels to be cut off from
the histogram of the input image. The value must be in range of [0.0, 50.0]. Default: ``0.0``.
ignore (Union[int, sequence], optional): The background pixel values to be ignored, each of
which must be in range of [0, 255]. Default: ``None``.
prob (float, optional): Probability of the image being automatically contrasted, which
must be in range of [0.0, 1.0]. Default: ``0.5``.
Raises:
TypeError: If `cutoff` is not of type float.
TypeError: If `ignore` is not of type integer or sequence of integer.
TypeError: If `prob` is not of type float.
ValueError: If `cutoff` is not in range [0.0, 50.0).
ValueError: If `ignore` is not in range [0, 255].
ValueError: If `prob` is not in range [0.0, 1.0].
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.RandomAutoContrast(cutoff=0.0, ignore=None, prob=0.5)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomAutoContrast(cutoff=0.0, ignore=None, prob=1.0)(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_random_auto_contrast
def __init__(self, cutoff=0.0, ignore=None, prob=0.5):
super().__init__()
if ignore is None:
ignore = []
if isinstance(ignore, int):
ignore = [ignore]
self.cutoff = cutoff
self.ignore = ignore
self.prob = prob
self.implementation = Implementation.C
def parse(self):
return cde.RandomAutoContrastOperation(self.cutoff, self.ignore, self.prob)
[docs]class RandomColor(ImageTensorOperation, PyTensorOperation):
"""
Adjust the color of the input image by a fixed or random degree.
This operation works only with 3-channel color images.
Args:
degrees (Sequence[float], optional): Range of random color adjustment degrees, which must be non-negative.
It should be in (min, max) format. If min=max, then it is a
single fixed magnitude operation. Default: ``(0.1, 1.9)``.
Raises:
TypeError: If `degrees` is not of type Sequence[float].
ValueError: If `degrees` is negative.
RuntimeError: If given tensor shape is not <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.RandomColor((0.5, 2.0))]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomColor((0.1, 1.9))(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_positive_degrees
def __init__(self, degrees=(0.1, 1.9)):
super().__init__()
self.degrees = degrees
def parse(self):
return cde.RandomColorOperation(*self.degrees)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be color adjusted.
Returns:
PIL Image, color adjusted image.
"""
return util.random_color(img, self.degrees)
[docs]class RandomColorAdjust(ImageTensorOperation, PyTensorOperation):
"""
Randomly adjust the brightness, contrast, saturation, and hue of the input image.
Note:
This operation is executed on the CPU by default, but it is also supported
to be executed on the GPU or Ascend via heterogeneous acceleration.
Args:
brightness (Union[float, Sequence[float]], optional): Brightness adjustment factor. Default: ``(1, 1)``.
Cannot be negative.
If it is a float, the factor is uniformly chosen from the range [max(0, 1-brightness), 1+brightness].
If it is a sequence, it should be [min, max] for the range.
contrast (Union[float, Sequence[float]], optional): Contrast adjustment factor. Default: ``(1, 1)``.
Cannot be negative.
If it is a float, the factor is uniformly chosen from the range [max(0, 1-contrast), 1+contrast].
If it is a sequence, it should be [min, max] for the range.
saturation (Union[float, Sequence[float]], optional): Saturation adjustment factor. Default: ``(1, 1)``.
Cannot be negative.
If it is a float, the factor is uniformly chosen from the range [max(0, 1-saturation), 1+saturation].
If it is a sequence, it should be [min, max] for the range.
hue (Union[float, Sequence[float]], optional): Hue adjustment factor. Default: ``(0, 0)``.
If it is a float, the range will be [-hue, hue]. Value should be 0 <= hue <= 0.5.
If it is a sequence, it should be [min, max] where -0.5 <= min <= max <= 0.5.
Raises:
TypeError: If `brightness` is not of type float or Sequence[float].
TypeError: If `contrast` is not of type float or Sequence[float].
TypeError: If `saturation` is not of type float or Sequence[float].
TypeError: If `hue` is not of type float or Sequence[float].
ValueError: If `brightness` is negative.
ValueError: If `contrast` is negative.
ValueError: If `saturation` is negative.
ValueError: If `hue` is not in range [-0.5, 0.5].
RuntimeError: If given tensor shape is not <H, W, C>.
Supported Platforms:
``CPU`` ``GPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transform_op = vision.RandomColorAdjust(brightness=(0.5, 1),
... contrast=(0.4, 1),
... saturation=(0.3, 1))
>>> transforms_list = [transform_op]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomColorAdjust(brightness=(0.5, 1), contrast=(0.4, 1), saturation=(0.3, 1))(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_random_color_adjust
def __init__(self, brightness=(1, 1), contrast=(1, 1), saturation=(1, 1), hue=(0, 0)):
super().__init__()
brightness = self.__expand_values(brightness)
contrast = self.__expand_values(contrast)
saturation = self.__expand_values(saturation)
hue = self.__expand_values(
hue, center=0, bound=(-0.5, 0.5), non_negative=False)
self.brightness = brightness
self.contrast = contrast
self.saturation = saturation
self.hue = hue
def parse(self):
return cde.RandomColorAdjustOperation(self.brightness, self.contrast, self.saturation, self.hue)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL image): Image to be randomly color adjusted.
Returns:
PIL Image, randomly color adjusted image.
"""
return util.random_color_adjust(img, self.brightness, self.contrast, self.saturation, self.hue)
def __expand_values(self, value, center=1, bound=(0, FLOAT_MAX_INTEGER), non_negative=True):
"""Expand input value for vision adjustment factor."""
if isinstance(value, numbers.Number):
value = [center - value, center + value]
if non_negative:
value[0] = max(0, value[0])
check_range(value, bound)
return (value[0], value[1])
[docs]class RandomCrop(ImageTensorOperation, PyTensorOperation):
"""
Crop the input image at a random location. If input image size is smaller than output size,
input image will be padded before cropping.
Note:
If the input image is more than one, then make sure that the image size is the same.
Args:
size (Union[int, Sequence[int]]): The output size of the cropped image. The size value(s) must be positive.
If size is an integer, a square crop of size (size, size) is returned.
If size is a sequence of length 2, an image of size (height, width) will be cropped.
padding (Union[int, Sequence[int]], optional): The number of pixels to pad each border of the image.
The padding value(s) must be non-negative. Default: ``None``.
If `padding` is not ``None``, pad image first with padding values.
If a single number is provided, pad all borders with this value.
If a tuple or lists of 2 values are provided, pad the (left and right)
with the first value and (top and bottom) with the second value.
If 4 values are provided as a list or tuple,
pad the left, top, right and bottom respectively.
pad_if_needed (bool, optional): Pad the image if either side is smaller than
the given output size. Default: ``False``.
fill_value (Union[int, tuple[int]], optional): The pixel intensity of the borders, only valid for
padding_mode Border.CONSTANT. 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``.
padding_mode (Border, optional): The method of padding. Default: ``Border.CONSTANT``. It can be any of
``Border.CONSTANT``, ``Border.EDGE``, ``Border.REFLECT``, ``Border.SYMMETRIC``.
- ``Border.CONSTANT`` , means it fills the border with constant values.
- ``Border.EDGE`` , means it pads with the last value on the edge.
- ``Border.REFLECT`` , means it reflects the values on the edge omitting the last
value of edge.
- ``Border.SYMMETRIC`` , means it reflects the values on the edge repeating the last
value of edge.
Raises:
TypeError: If `size` is not of type int or Sequence[int].
TypeError: If `padding` is not of type int or Sequence[int].
TypeError: If `pad_if_needed` is not of type boolean.
TypeError: If `fill_value` is not of type int or tuple[int].
TypeError: If `padding_mode` is not of type :class:`mindspore.dataset.vision.Border` .
ValueError: If `size` is not positive.
ValueError: If `padding` is negative.
ValueError: If `fill_value` is not in range [0, 255].
RuntimeError: If given tensor shape is not <H, W> or <..., H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Border
>>>
>>> # Use the transform in dataset pipeline mode
>>> random_crop_op = vision.RandomCrop(64, [16, 16, 16, 16], padding_mode=Border.EDGE)
>>> transforms_list = [random_crop_op]
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(64, 64, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomCrop(8, [10, 10, 10, 10], padding_mode=Border.EDGE)(data)
>>> print(output.shape, output.dtype)
(8, 8, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_random_crop
def __init__(self, size, padding=None, pad_if_needed=False, fill_value=0, padding_mode=Border.CONSTANT):
super().__init__()
if isinstance(size, int):
size = (size, size)
if padding is None:
padding = (0, 0, 0, 0)
else:
padding = parse_padding(padding)
if isinstance(fill_value, int):
fill_value = tuple([fill_value] * 3)
self.size = size
self.padding = padding
self.pad_if_needed = pad_if_needed
self.fill_value = fill_value
self.padding_mode = padding_mode
def parse(self):
return cde.RandomCropOperation(self.size, self.padding, self.pad_if_needed, self.fill_value,
Border.to_c_type(self.padding_mode))
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be randomly cropped.
Returns:
PIL Image, cropped image.
"""
return util.random_crop(img, self.size, self.padding, self.pad_if_needed,
self.fill_value, Border.to_python_type(self.padding_mode))
[docs]class RandomCropDecodeResize(ImageTensorOperation):
"""
A combination of `Crop` , `Decode` and `Resize` . It will get better performance for JPEG images. This operation
will crop the input image at a random location, decode the cropped image in RGB mode, and resize the decoded image.
Args:
size (Union[int, Sequence[int]]): The output size of the resized image. The size value(s) must be positive.
If size is an integer, a square crop of size (size, size) is returned.
If size is a sequence of length 2, it should be (height, width).
scale (Union[list, tuple], optional): Range [min, max) of respective size of the
original size to be cropped, which must be non-negative. Default: ``(0.08, 1.0)``.
ratio (Union[list, tuple], optional): Range [min, max) of aspect ratio to be
cropped, which must be non-negative. Default: ``(3. / 4., 4. / 3.)``.
interpolation (Inter, optional): Image interpolation method defined by :class:`~.vision.Inter` .
Default: ``Inter.BILINEAR``.
max_attempts (int, optional): The maximum number of attempts to propose a valid crop_area. Default: ``10``.
If exceeded, fall back to use center_crop instead. The `max_attempts` value must be positive.
Raises:
TypeError: If `size` is not of type int or Sequence[int].
TypeError: If `scale` is not of type tuple.
TypeError: If `ratio` is not of type tuple.
TypeError: If `interpolation` is not of type :class:`~.vision.Inter` .
TypeError: If `max_attempts` is not of type integer.
ValueError: If `size` is not positive.
ValueError: If `scale` is negative.
ValueError: If `ratio` is negative.
ValueError: If `max_attempts` is not positive.
RuntimeError: If given tensor is not a 1D sequence.
Supported Platforms:
``CPU``
Examples:
>>> import os
>>> import numpy as np
>>> from PIL import Image, ImageDraw
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> class MyDataset:
... def __init__(self):
... self.data = []
... img = Image.new("RGB", (300, 300), (255, 255, 255))
... draw = ImageDraw.Draw(img)
... draw.ellipse(((0, 0), (100, 100)), fill=(255, 0, 0), outline=(255, 0, 0), width=5)
... img.save("./1.jpg")
... data = np.fromfile("./1.jpg", np.uint8)
... self.data.append(data)
...
... def __getitem__(self, index):
... return self.data[0]
...
... def __len__(self):
... return 5
>>>
>>> my_dataset = MyDataset()
>>> generator_dataset = ds.GeneratorDataset(my_dataset, column_names="image")
>>> resize_crop_decode_op = vision.RandomCropDecodeResize(size=(50, 75),
... scale=(0.25, 0.5),
... interpolation=Inter.NEAREST,
... max_attempts=5)
>>> transforms_list = [resize_crop_decode_op]
>>> generator_dataset = generator_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in generator_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(50, 75, 3) uint8
>>> os.remove("./1.jpg")
>>>
>>> # Use the transform in eager mode
>>> img = Image.new("RGB", (300, 300), (255, 255, 255))
>>> draw = ImageDraw.Draw(img)
>>> draw.polygon([(50, 50), (150, 50), (100, 150)], fill=(0, 255, 0), outline=(0, 255, 0))
>>> img.save("./2.jpg")
>>> data = np.fromfile("./2.jpg", np.uint8)
>>> output = vision.RandomCropDecodeResize(size=(50, 75), scale=(0, 10.0), ratio=(0.5, 0.5),
... interpolation=Inter.BILINEAR, max_attempts=1)(data)
>>> print(np.array(output).shape, np.array(output).dtype)
(50, 75, 3) uint8
>>> os.remove("./2.jpg")
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_random_resize_crop
def __init__(self, size, scale=(0.08, 1.0), ratio=(3. / 4., 4. / 3.),
interpolation=Inter.BILINEAR, max_attempts=10):
super().__init__()
if isinstance(size, int):
size = (size, size)
self.size = size
self.scale = scale
self.ratio = ratio
self.interpolation = interpolation
self.max_attempts = max_attempts
self.implementation = Implementation.C
def __call__(self, img):
if not isinstance(img, np.ndarray):
raise TypeError(
"Input should be an encoded image in 1-D NumPy format, got {}.".format(type(img)))
if img.ndim != 1 or img.dtype.type is not np.uint8:
raise TypeError("Input should be an encoded image with uint8 type in 1-D NumPy format, " +
"got format:{}, dtype:{}.".format(type(img), img.dtype.type))
return super().__call__(img)
def parse(self):
return cde.RandomCropDecodeResizeOperation(self.size, self.scale, self.ratio,
Inter.to_c_type(self.interpolation),
self.max_attempts)
[docs]class RandomCropWithBBox(ImageTensorOperation):
"""
Crop the input image at a random location and adjust bounding boxes accordingly.
Args:
size (Union[int, Sequence[int]]): The output size of the cropped image. The size value(s) must be positive.
If size is an integer, a square crop of size (size, size) is returned.
If size is a sequence of length 2, an image of size (height, width) will be cropped.
padding (Union[int, Sequence[int]], optional): The number of pixels to pad the image
The padding value(s) must be non-negative. Default: ``None``.
If `padding` is not ``None``, first pad image with padding values.
If a single number is provided, pad all borders with this value.
If a tuple or lists of 2 values are provided, pad the (left and right)
with the first value and (top and bottom) with the second value.
If 4 values are provided as a list or tuple, pad the left, top, right and bottom respectively.
pad_if_needed (bool, optional): Pad the image if either side is smaller than
the given output size. Default: ``False``.
fill_value (Union[int, tuple[int]], optional): The pixel intensity of the borders, only valid for
padding_mode Border.CONSTANT. 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``.
padding_mode (Border, optional): The method of padding. Default: ``Border.CONSTANT``. It can be any of
``Border.CONSTANT``, ``Border.EDGE``, ``Border.REFLECT``, ``Border.SYMMETRIC``.
- ``Border.CONSTANT`` , means it fills the border with constant values.
- ``Border.EDGE`` , means it pads with the last value on the edge.
- ``Border.REFLECT`` , means it reflects the values on the edge omitting the last
value of edge.
- ``Border.SYMMETRIC`` , means it reflects the values on the edge repeating the last
value of edge.
Raises:
TypeError: If `size` is not of type int or Sequence[int].
TypeError: If `padding` is not of type int or Sequence[int].
TypeError: If `pad_if_needed` is not of type boolean.
TypeError: If `fill_value` is not of type int or tuple[int].
TypeError: If `padding_mode` is not of type :class:`mindspore.dataset.vision.Border` .
ValueError: If `size` is not positive.
ValueError: If `padding` is negative.
ValueError: If `fill_value` is not in range [0, 255].
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.float32)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> func = lambda img: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(np.float32))
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=[func],
... input_columns=["image"],
... output_columns=["image", "bbox"])
>>> random_crop_with_bbox_op = vision.RandomCropWithBBox([64, 64], [20, 20, 20, 20])
>>> transforms_list = [random_crop_with_bbox_op]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image", "bbox"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... print(item["bbox"].shape, item["bbox"].dtype)
... break
(64, 64, 3) float32
(1, 4) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.float32)
>>> func = lambda img: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(data.dtype))
>>> func_data, func_bboxes = func(data)
>>> output = vision.RandomCropWithBBox([64, 64], [20, 20, 20, 20])(func_data, func_bboxes)
>>> print(output[0].shape, output[0].dtype)
(64, 64, 3) float32
>>> print(output[1].shape, output[1].dtype)
(1, 4) float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_random_crop
def __init__(self, size, padding=None, pad_if_needed=False, fill_value=0, padding_mode=Border.CONSTANT):
super().__init__()
if isinstance(size, int):
size = (size, size)
if padding is None:
padding = (0, 0, 0, 0)
else:
padding = parse_padding(padding)
if isinstance(fill_value, int):
fill_value = tuple([fill_value] * 3)
self.size = size
self.padding = padding
self.pad_if_needed = pad_if_needed
self.fill_value = fill_value
self.padding_mode = padding_mode
self.implementation = Implementation.C
def parse(self):
border_type = Border.to_c_type(self.padding_mode)
return cde.RandomCropWithBBoxOperation(self.size, self.padding, self.pad_if_needed, self.fill_value,
border_type)
[docs]class RandomEqualize(ImageTensorOperation):
"""
Apply histogram equalization on the input image with a given probability.
Args:
prob (float, optional): Probability of the image being equalized, which
must be in range of [0.0, 1.0]. Default: ``0.5``.
Raises:
TypeError: If `prob` is not of type float.
ValueError: If `prob` is not in range [0.0, 1.0].
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.RandomEqualize(0.5)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomEqualize(1.0)(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_prob
def __init__(self, prob=0.5):
super().__init__()
self.prob = prob
self.implementation = Implementation.C
def parse(self):
return cde.RandomEqualizeOperation(self.prob)
[docs]class RandomErasing(PyTensorOperation):
"""
Randomly erase pixels within a random selected rectangle erea on the input numpy.ndarray image.
See `Random Erasing Data Augmentation <https://arxiv.org/pdf/1708.04896.pdf>`_ .
Args:
prob (float, optional): Probability of performing erasing, which
must be in range of [0.0, 1.0]. Default: ``0.5``.
scale (Sequence[float, float], optional): Range of area scale of the erased area relative
to the original image to select from, arranged in order of (min, max).
Default: ``(0.02, 0.33)``.
ratio (Sequence[float, float], optional): Range of aspect ratio of the erased area to select
from, arraged in order of (min, max). Default: ``(0.3, 3.3)``.
value (Union[int, str, Sequence[int, int, int]]): Pixel value used to pad the erased area.
If a single integer is provided, it will be used for all RGB channels.
If a sequence of length 3 is provided, it will be used for R, G, B channels respectively.
If a string of ``'random'`` is provided, each pixel will be erased with a random value obtained
from a standard normal distribution. Default: ``0``.
inplace (bool, optional): Whether to apply erasing inplace. Default: ``False``.
max_attempts (int, optional): The maximum number of attempts to propose a valid
erased area, beyond which the original image will be returned. Default: ``10``.
Raises:
TypeError: If `prob` is not of type float.
TypeError: If `scale` is not of type sequence.
TypeError: If `ratio` is not of type sequence.
TypeError: If `value` is not of type integer, string, or sequence.
TypeError: If `inplace` is not of type boolean.
TypeError: If `max_attempts` is not of type integer.
ValueError: If `prob` is not in range of [0.0, 1.0].
ValueError: If `scale` is negative.
ValueError: If `ratio` is negative.
ValueError: If `value` is not in range of [0, 255].
ValueError: If `max_attempts` is not positive.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.transforms import Compose
>>>
>>> # Use the transform in dataset pipeline mode
>>> transforms_list = Compose([vision.ToTensor(),
... vision.RandomErasing(value='random')])
>>> # apply the transform to dataset through map function
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns="image")
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(3, 100, 100) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(254, 255, size=(3, 100, 100)).astype(np.uint8)
>>> output = vision.RandomErasing(prob=1.0, max_attempts=1)(data)
>>> print(output.shape, output.dtype)
(3, 100, 100) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_random_erasing
def __init__(self, prob=0.5, scale=(0.02, 0.33), ratio=(0.3, 3.3), value=0, inplace=False, max_attempts=10):
super().__init__()
self.prob = prob
self.scale = scale
self.ratio = ratio
self.value = value
self.inplace = inplace
self.max_attempts = max_attempts
self.implementation = Implementation.PY
def _execute_py(self, np_img):
"""
Execute method.
Args:
np_img (numpy.ndarray): image in shape of <C, H, W> to be randomly erased.
Returns:
numpy.ndarray, erased image.
"""
bounded = True
if self.prob > random.random():
i, j, erase_h, erase_w, erase_value = util.get_erase_params(np_img, self.scale, self.ratio,
self.value, bounded, self.max_attempts)
return util.erase(np_img, i, j, erase_h, erase_w, erase_value, self.inplace)
return np_img
[docs]class RandomGrayscale(PyTensorOperation):
"""
Randomly convert the input PIL Image to grayscale.
Args:
prob (float, optional): Probability of performing grayscale conversion,
which must be in range of [0.0, 1.0]. Default: ``0.1``.
Raises:
TypeError: If `prob` is not of type float.
ValueError: If `prob` is not in range of [0.0, 1.0].
Supported Platforms:
``CPU``
Examples:
>>> import os
>>> import numpy as np
>>> from PIL import Image, ImageDraw
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.transforms import Compose
>>>
>>> # Use the transform in dataset pipeline mode
>>> class MyDataset:
... def __init__(self):
... self.data = []
... img = Image.new("RGB", (300, 300), (255, 255, 255))
... draw = ImageDraw.Draw(img)
... draw.ellipse(((0, 0), (100, 100)), fill=(255, 0, 0), outline=(255, 0, 0), width=5)
... img.save("./1.jpg")
... data = np.fromfile("./1.jpg", np.uint8)
... self.data.append(data)
...
... def __getitem__(self, index):
... return self.data[0]
...
... def __len__(self):
... return 5
>>>
>>> my_dataset = MyDataset()
>>> generator_dataset = ds.GeneratorDataset(my_dataset, column_names="image")
>>> transforms_list = Compose([vision.Decode(to_pil=True),
... vision.RandomGrayscale(0.3),
... vision.ToTensor()])
>>> # apply the transform to dataset through map function
>>> generator_dataset = generator_dataset.map(operations=transforms_list, input_columns="image")
>>> for item in generator_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(3, 300, 300) float32
>>> os.remove("./1.jpg")
>>>
>>> # Use the transform in eager mode
>>> img = Image.new("RGB", (300, 300), (255, 255, 255))
>>> draw = ImageDraw.Draw(img)
>>> draw.polygon([(50, 50), (150, 50), (100, 150)], fill=(0, 255, 0), outline=(0, 255, 0))
>>> img.save("./2.jpg")
>>> data = Image.open("./2.jpg")
>>> output = vision.RandomGrayscale(1.0)(data)
>>> print(np.array(output).shape, np.array(output).dtype)
(300, 300, 3) uint8
>>> os.remove("./2.jpg")
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_prob
def __init__(self, prob=0.1):
super().__init__()
self.prob = prob
self.implementation = Implementation.PY
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be randomly converted to grayscale.
Returns:
PIL Image, randomly converted grayscale image, which has the same number of channels as the input image.
If input image has 1 channel, the output grayscale image will have 1 channel.
If input image has 3 channels, the output grayscale image will have 3 identical channels.
"""
if img.mode == 'L':
num_output_channels = 1
else:
num_output_channels = 3
if self.prob > random.random():
return util.grayscale(img, num_output_channels=num_output_channels)
return img
[docs]class RandomHorizontalFlip(ImageTensorOperation, PyTensorOperation):
"""
Randomly flip the input image horizontally with a given probability.
Args:
prob (float, optional): Probability of the image being flipped,
which must be in range of [0.0, 1.0]. Default: ``0.5``.
Raises:
TypeError: If `prob` is not of type float.
ValueError: If `prob` is not in range [0.0, 1.0].
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.RandomHorizontalFlip(0.75)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomHorizontalFlip(1.0)(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_prob
def __init__(self, prob=0.5):
super().__init__()
self.prob = prob
def parse(self):
return cde.RandomHorizontalFlipOperation(self.prob)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be horizontally flipped.
Returns:
PIL Image, randomly horizontally flipped image.
"""
return util.random_horizontal_flip(img, self.prob)
[docs]class RandomHorizontalFlipWithBBox(ImageTensorOperation):
"""
Randomly flip the input image and its bounding box horizontally with a given probability.
Args:
prob (float, optional): Probability of the image being flipped,
which must be in range of [0.0, 1.0]. Default: ``0.5``.
Raises:
TypeError: If `prob` is not of type float.
ValueError: If `prob` is not in range [0.0, 1.0].
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.float32)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> func = lambda img: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(np.float32))
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=[func],
... input_columns=["image"],
... output_columns=["image", "bbox"])
>>> transforms_list = [vision.RandomHorizontalFlipWithBBox(0.70)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list,
... input_columns=["image", "bbox"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... print(item["bbox"].shape, item["bbox"].dtype)
... break
(100, 100, 3) float32
(1, 4) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.float32)
>>> func = lambda img: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(data.dtype))
>>> func_data, func_bboxes = func(data)
>>> output = vision.RandomHorizontalFlipWithBBox(1)(func_data, func_bboxes)
>>> print(output[0].shape, output[0].dtype)
(100, 100, 3) float32
>>> print(output[1].shape, output[1].dtype)
(1, 4) float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_prob
def __init__(self, prob=0.5):
super().__init__()
self.prob = prob
self.implementation = Implementation.C
def parse(self):
return cde.RandomHorizontalFlipWithBBoxOperation(self.prob)
[docs]class RandomInvert(ImageTensorOperation):
"""
Randomly invert the colors of image with a given probability.
Args:
prob (float, optional): Probability of the image being inverted,
which must be in range of [0.0, 1.0]. Default: ``0.5``.
Raises:
TypeError: If `prob` is not of type float.
ValueError: If `prob` is not in range [0.0, 1.0].
RuntimeError: If given tensor shape is not <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.RandomInvert(0.5)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomInvert(1.0)(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_prob
def __init__(self, prob=0.5):
super().__init__()
self.prob = prob
self.implementation = Implementation.C
def parse(self):
return cde.RandomInvertOperation(self.prob)
[docs]class RandomLighting(ImageTensorOperation, PyTensorOperation):
"""
Add AlexNet-style PCA-based noise to an image. The eigenvalue and eigenvectors for Alexnet's PCA noise is
calculated from the imagenet dataset.
Args:
alpha (float, optional): Intensity of the image, which must be non-negative. Default: ``0.05``.
Raises:
TypeError: If `alpha` is not of type float.
ValueError: If `alpha` is negative.
RuntimeError: If given tensor shape is not <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.RandomLighting(0.1)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomLighting(0.1)(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_alpha
def __init__(self, alpha=0.05):
super().__init__()
self.alpha = alpha
def parse(self):
return cde.RandomLightingOperation(self.alpha)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be added AlexNet-style PCA-based noise.
Returns:
PIL Image, image with noise added.
"""
return util.random_lighting(img, self.alpha)
[docs]class RandomPerspective(PyTensorOperation):
"""
Randomly apply perspective transformation to the input PIL Image with a given probability.
Args:
distortion_scale (float, optional): Scale of distortion, in range of [0.0, 1.0]. Default: ``0.5``.
prob (float, optional): Probability of performing perspective transformation, which
must be in range of [0.0, 1.0]. Default: ``0.5``.
interpolation (Inter, optional): Image interpolation method defined by :class:`~.vision.Inter` .
Default: ``Inter.BICUBIC``.
Raises:
TypeError: If `distortion_scale` is not of type float.
TypeError: If `prob` is not of type float.
TypeError: If `interpolation` is not of type :class:`~.vision.Inter` .
ValueError: If `distortion_scale` is not in range of [0.0, 1.0].
ValueError: If `prob` is not in range of [0.0, 1.0].
Supported Platforms:
``CPU``
Examples:
>>> import os
>>> import numpy as np
>>> from PIL import Image, ImageDraw
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.transforms import Compose
>>>
>>> # Use the transform in dataset pipeline mode
>>> class MyDataset:
... def __init__(self):
... self.data = []
... img = Image.new("RGB", (300, 300), (255, 255, 255))
... draw = ImageDraw.Draw(img)
... draw.ellipse(((0, 0), (100, 100)), fill=(255, 0, 0), outline=(255, 0, 0), width=5)
... img.save("./1.jpg")
... data = np.fromfile("./1.jpg", np.uint8)
... self.data.append(data)
...
... def __getitem__(self, index):
... return self.data[0]
...
... def __len__(self):
... return 5
>>>
>>> my_dataset = MyDataset()
>>> generator_dataset = ds.GeneratorDataset(my_dataset, column_names="image")
>>> transforms_list = Compose([vision.Decode(to_pil=True),
... vision.RandomPerspective(prob=0.1),
... vision.ToTensor()])
>>> # apply the transform to dataset through map function
>>> generator_dataset = generator_dataset.map(operations=transforms_list, input_columns="image")
>>> for item in generator_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(3, 300, 300) float32
>>> os.remove("./1.jpg")
>>>
>>> # Use the transform in eager mode
>>> img = Image.new("RGB", (300, 300), (255, 255, 255))
>>> draw = ImageDraw.Draw(img)
>>> draw.polygon([(50, 50), (150, 50), (100, 150)], fill=(0, 255, 0), outline=(0, 255, 0))
>>> img.save("./2.jpg")
>>> data = Image.open("./2.jpg")
>>> output = vision.RandomPerspective(prob=1.0)(data)
>>> print(np.array(output).shape, np.array(output).dtype)
(300, 300, 3) uint8
>>> os.remove("./2.jpg")
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_random_perspective
def __init__(self, distortion_scale=0.5, prob=0.5, interpolation=Inter.BICUBIC):
super().__init__()
self.distortion_scale = distortion_scale
self.prob = prob
self.interpolation = interpolation
self.implementation = Implementation.PY
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be applied randomly perspective transformation.
Returns:
PIL Image, image applied randomly perspective transformation.
"""
if not is_pil(img):
raise ValueError("Input image should be a Pillow image.")
if self.prob > random.random():
start_points, end_points = util.get_perspective_params(
img, self.distortion_scale)
return util.perspective(img, start_points, end_points, Inter.to_python_type(self.interpolation))
return img
[docs]class RandomPosterize(ImageTensorOperation):
"""
Reduce the bit depth of the color channels of image with a given probability
to create a high contrast and vivid color image.
Reduce the number of bits for each color channel to posterize the input image randomly with a given probability.
Args:
bits (Union[int, Sequence[int]], optional): Range of random posterize to compress image.
Bits values must be in range of [1,8], and include at
least one integer value in the given range. It must be in
(min, max) or integer format. If min=max, then it is a single fixed
magnitude operation. Default: ``(8, 8)``.
Raises:
TypeError: If `bits` is not of type integer or sequence of integer.
ValueError: If `bits` is not in range [1, 8].
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.RandomPosterize((6, 8))]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomPosterize(1)(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_random_posterize
def __init__(self, bits=(8, 8)):
super().__init__()
self.bits = bits
self.implementation = Implementation.C
def parse(self):
bits = self.bits
if isinstance(bits, int):
bits = (bits, bits)
return cde.RandomPosterizeOperation(bits)
[docs]class RandomResizedCrop(ImageTensorOperation, PyTensorOperation):
"""
This operation will crop the input image randomly,
and resize the cropped image using a selected interpolation mode :class:`~.vision.Inter` .
Note:
If the input image is more than one, then make sure that the image size is the same.
Args:
size (Union[int, Sequence[int]]): The output size of the resized image. The size value(s) must be positive.
If size is an integer, a square of size (size, size) will be cropped with this value.
If size is a sequence of length 2, an image of size (height, width) will be cropped.
scale (Union[list, tuple], optional): Range [min, max) of respective size of the original
size to be cropped, which must be non-negative. Default: ``(0.08, 1.0)``.
ratio (Union[list, tuple], optional): Range [min, max) of aspect ratio to be
cropped, which must be non-negative. Default: ``(3. / 4., 4. / 3.)``.
interpolation (Inter, optional): Image interpolation method defined by :class:`~.vision.Inter` .
Default: ``Inter.BILINEAR``.
max_attempts (int, optional): The maximum number of attempts to propose a valid
crop_area. Default: ``10``. If exceeded, fall back to use center_crop instead.
Raises:
TypeError: If `size` is not of type int or Sequence[int].
TypeError: If `scale` is not of type tuple or list.
TypeError: If `ratio` is not of type tuple or list.
TypeError: If `interpolation` is not of type :class:`~.vision.Inter` .
TypeError: If `max_attempts` is not of type int.
ValueError: If `size` is not positive.
ValueError: If `scale` is negative.
ValueError: If `ratio` is negative.
ValueError: If `max_attempts` is not positive.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> resize_crop_op = vision.RandomResizedCrop(size=(50, 75), scale=(0.25, 0.5),
... interpolation=Inter.BILINEAR)
>>> transforms_list = [resize_crop_op]
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(50, 75, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomResizedCrop(size=(50, 75), scale=(0.25, 0.5), interpolation=Inter.BILINEAR)(data)
>>> print(output.shape, output.dtype)
(50, 75, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_random_resize_crop
def __init__(self, size, scale=(0.08, 1.0), ratio=(3. / 4., 4. / 3.),
interpolation=Inter.BILINEAR, max_attempts=10):
super().__init__()
if isinstance(size, int):
size = (size, size)
self.size = size
self.scale = scale
self.ratio = ratio
self.interpolation = interpolation
if interpolation in [Inter.AREA, Inter.PILCUBIC]:
self.implementation = Implementation.C
elif interpolation == Inter.ANTIALIAS:
self.implementation = Implementation.PY
self.max_attempts = max_attempts
def parse(self):
if self.interpolation == Inter.ANTIALIAS:
raise TypeError("Current Interpolation is not supported with NumPy input.")
return cde.RandomResizedCropOperation(self.size, self.scale, self.ratio, Inter.to_c_type(self.interpolation),
self.max_attempts)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be randomly cropped and resized.
Returns:
PIL Image, randomly cropped and resized image.
"""
if self.interpolation in [Inter.AREA, Inter.PILCUBIC]:
raise TypeError("Current Interpolation is not supported with PIL input.")
return util.random_resize_crop(img, self.size, self.scale, self.ratio,
Inter.to_python_type(self.interpolation), self.max_attempts)
[docs]class RandomResizedCropWithBBox(ImageTensorOperation):
"""
Crop the input image to a random size and aspect ratio and adjust bounding boxes accordingly.
Args:
size (Union[int, Sequence[int]]): The size of the output image. The size value(s) must be positive.
If size is an integer, a square crop of size (size, size) is returned.
If size is a sequence of length 2, it should be (height, width).
scale (Union[list, tuple], optional): Range (min, max) of respective size of the original
size to be cropped, which must be non-negative. Default: ``(0.08, 1.0)``.
ratio (Union[list, tuple], optional): Range (min, max) of aspect ratio to be
cropped, which must be non-negative. Default: ``(3. / 4., 4. / 3.)``.
interpolation (Inter, optional): Image interpolation method defined by :class:`~.vision.Inter` .
Default: ``Inter.BILINEAR``.
max_attempts (int, optional): The maximum number of attempts to propose a valid
crop area. Default: ``10``. If exceeded, fall back to use center crop instead.
Raises:
TypeError: If `size` is not of type int or Sequence[int].
TypeError: If `scale` is not of type tuple.
TypeError: If `ratio` is not of type tuple.
TypeError: If `interpolation` is not of type Inter.
TypeError: If `max_attempts` is not of type integer.
ValueError: If `size` is not positive.
ValueError: If `scale` is negative.
ValueError: If `ratio` is negative.
ValueError: If `max_attempts` is not positive.
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.float32)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> func = lambda img: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(np.float32))
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=[func],
... input_columns=["image"],
... output_columns=["image", "bbox"])
>>> bbox_op = vision.RandomResizedCropWithBBox(size=50, interpolation=Inter.NEAREST)
>>> transforms_list = [bbox_op]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list,
... input_columns=["image", "bbox"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... print(item["bbox"].shape, item["bbox"].dtype)
... break
(50, 50, 3) float32
(1, 4) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.float32)
>>> func = lambda img: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(data.dtype))
>>> func_data, func_bboxes = func(data)
>>> output = vision.RandomResizedCropWithBBox((16, 64), (0.5, 0.5), (0.5, 0.5))(func_data, func_bboxes)
>>> print(output[0].shape, output[0].dtype)
(16, 64, 3) float32
>>> print(output[1].shape, output[1].dtype)
(1, 4) float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_random_resize_crop
def __init__(self, size, scale=(0.08, 1.0), ratio=(3. / 4., 4. / 3.),
interpolation=Inter.BILINEAR, max_attempts=10):
super().__init__()
if isinstance(size, int):
size = (size, size)
self.size = size
self.scale = scale
self.ratio = ratio
self.interpolation = interpolation
self.max_attempts = max_attempts
self.implementation = Implementation.C
def parse(self):
return cde.RandomResizedCropWithBBoxOperation(self.size, self.scale, self.ratio,
Inter.to_c_type(self.interpolation), self.max_attempts)
[docs]class RandomResize(ImageTensorOperation):
"""
Resize the input image using :class:`~.vision.Inter` , a randomly selected interpolation mode.
Args:
size (Union[int, Sequence[int]]): The output size of the resized image. The size value(s) must be positive.
If size is an integer, smaller edge of the image will be resized to this value with
the same image aspect ratio.
If size is a sequence of length 2, it should be (height, width).
Raises:
TypeError: If `size` is not of type int or Sequence[int].
ValueError: If `size` is not positive.
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> # 1) randomly resize image, keeping aspect ratio
>>> transforms_list1 = [vision.RandomResize(50)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list1, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(50, 50, 3) uint8
>>> # 2) randomly resize image to landscape style
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list2 = [vision.RandomResize((40, 60))]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list2, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(40, 60, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomResize(10)(data)
>>> print(output.shape, output.dtype)
(10, 10, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_resize
def __init__(self, size):
super().__init__()
self.size = size
self.implementation = Implementation.C
def parse(self):
size = self.size
if isinstance(size, int):
size = (size,)
return cde.RandomResizeOperation(size)
[docs]class RandomResizeWithBBox(ImageTensorOperation):
"""
Tensor operation to resize the input image
using a randomly selected interpolation mode :class:`~.vision.Inter` and adjust
bounding boxes accordingly.
Args:
size (Union[int, Sequence[int]]): The output size of the resized image. The size value(s) must be positive.
If size is an integer, smaller edge of the image will be resized to this value with
the same image aspect ratio.
If size is a sequence of length 2, it should be (height, width).
Raises:
TypeError: If `size` is not of type int or Sequence[int].
ValueError: If `size` is not positive.
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import copy
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.float32)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> func = lambda img: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(np.float32))
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=[func],
... input_columns=["image"],
... output_columns=["image", "bbox"])
>>> numpy_slices_dataset2 = copy.deepcopy(numpy_slices_dataset)
>>>
>>> # 1) randomly resize image with bounding boxes, keeping aspect ratio
>>> transforms_list1 = [vision.RandomResizeWithBBox(60)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list1,
... input_columns=["image", "bbox"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... print(item["bbox"].shape, item["bbox"].dtype)
... break
(60, 60, 3) float32
(1, 4) float32
>>>
>>> # 2) randomly resize image with bounding boxes to portrait style
>>> transforms_list2 = [vision.RandomResizeWithBBox((80, 60))]
>>> numpy_slices_dataset2 = numpy_slices_dataset2.map(operations=transforms_list2,
... input_columns=["image", "bbox"])
>>> for item in numpy_slices_dataset2.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... print(item["bbox"].shape, item["bbox"].dtype)
... break
(80, 60, 3) float32
(1, 4) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.float32)
>>> func = lambda img: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(data.dtype))
>>> func_data, func_bboxes = func(data)
>>> output = vision.RandomResizeWithBBox(64)(func_data, func_bboxes)
>>> print(output[0].shape, output[0].dtype)
(64, 64, 3) float32
>>> print(output[1].shape, output[1].dtype)
(1, 4) float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_resize
def __init__(self, size):
super().__init__()
self.size = size
self.implementation = Implementation.C
def parse(self):
size = self.size
if isinstance(size, int):
size = (size,)
return cde.RandomResizeWithBBoxOperation(size)
[docs]class RandomRotation(ImageTensorOperation, PyTensorOperation):
"""
Rotate the input image randomly within a specified range of degrees.
Args:
degrees (Union[int, float, sequence]): Range of random rotation degrees.
If `degrees` is a number, the range will be converted to (-degrees, degrees).
If `degrees` is a sequence, it should be (min, max).
resample (Inter, optional): Image interpolation method defined by :class:`~.vision.Inter` .
Default: ``Inter.NEAREST``.
expand (bool, optional): Optional expansion flag. Default: ``False``. If set to ``True``,
expand the output image to make it large enough to hold the entire rotated image.
If set to ``False`` or omitted, make the output image the same size as the input.
Note that the expand flag assumes rotation around the center and no translation.
center (tuple, optional): Optional center of rotation (a 2-tuple). Default: ``None``.
Origin is the top left corner. ``None`` sets to the center of the image.
fill_value (Union[int, tuple[int]], optional): Optional fill color for the area outside the rotated image.
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 `degrees` is not of type integer, float or sequence.
TypeError: If `resample` is not of type Inter.
TypeError: If `expand` is not of type boolean.
TypeError: If `center` is not of type tuple.
TypeError: If `fill_value` is not of type int or tuple[int].
ValueError: If `fill_value` is not in range [0, 255].
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> seed = ds.config.get_seed()
>>> ds.config.set_seed(12345)
>>> transforms_list = [vision.RandomRotation(degrees=5.0, resample=Inter.NEAREST, expand=True)]
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(107, 107, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomRotation(degrees=90, resample=Inter.NEAREST, expand=True)(data)
>>> print(output.shape, output.dtype)
(119, 119, 3) uint8
>>> ds.config.set_seed(seed)
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_random_rotation
def __init__(self, degrees, resample=Inter.NEAREST, expand=False, center=None, fill_value=0):
super().__init__()
if isinstance(degrees, (int, float)):
degrees = degrees % 360
degrees = [-degrees, degrees]
elif isinstance(degrees, (list, tuple)):
if degrees[1] - degrees[0] >= 360:
degrees = [-180, 180]
else:
degrees = [degrees[0] % 360, degrees[1] % 360]
if degrees[0] > degrees[1]:
degrees[1] += 360
if isinstance(fill_value, int):
fill_value = tuple([fill_value] * 3)
self.degrees = degrees
self.resample = resample
if resample in [Inter.AREA, Inter.PILCUBIC]:
self.implementation = Implementation.C
elif resample == Inter.ANTIALIAS:
self.implementation = Implementation.PY
self.expand = expand
self.py_center = center
self.c_center = center
if center is None:
self.c_center = ()
self.fill_value = fill_value
def parse(self):
if self.resample == Inter.ANTIALIAS:
raise TypeError("Current Interpolation is not supported with NumPy input.")
return cde.RandomRotationOperation(self.degrees, Inter.to_c_type(self.resample), self.expand, self.c_center,
self.fill_value)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be randomly rotated.
Returns:
PIL Image, randomly rotated image.
"""
if self.resample in [Inter.AREA, Inter.PILCUBIC]:
raise TypeError("Current Interpolation is not supported with PIL input.")
return util.random_rotation(img, self.degrees, Inter.to_python_type(self.resample), self.expand,
self.py_center, self.fill_value)
[docs]class RandomSelectSubpolicy(ImageTensorOperation):
"""
Choose a random sub-policy from a policy list to be applied on the input image.
Args:
policy (list[list[tuple[TensorOperation, float]]]): List of sub-policies to choose from.
A sub-policy is a list of tuple[operation, prob], where operation is a data processing operation and prob
is the probability that this operation will be applied, and the prob values must be in range [0.0, 1.0].
Once a sub-policy is selected, each operation within the sub-policy with be applied in sequence according
to its probability.
Raises:
TypeError: If `policy` contains invalid data processing operations.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> policy = [[(vision.RandomRotation((45, 45)), 0.5),
... (vision.RandomVerticalFlip(), 1),
... (vision.RandomColorAdjust(), 0.8)],
... [(vision.RandomRotation((90, 90)), 1),
... (vision.RandomColorAdjust(), 0.2)]]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=vision.RandomSelectSubpolicy(policy),
... input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> policy = [[(vision.RandomRotation((90, 90)), 1), (vision.RandomColorAdjust(), 1)]]
>>> output = vision.RandomSelectSubpolicy(policy)(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_random_select_subpolicy_op
def __init__(self, policy):
super().__init__()
self.policy = policy
self.implementation = Implementation.C
def parse(self):
policy = []
for list_one in self.policy:
policy_one = []
for list_two in list_one:
if list_two[0] and getattr(list_two[0], 'parse', None):
policy_one.append((list_two[0].parse(), list_two[1]))
else:
policy_one.append((list_two[0], list_two[1]))
policy.append(policy_one)
return cde.RandomSelectSubpolicyOperation(policy)
[docs]class RandomSharpness(ImageTensorOperation, PyTensorOperation):
"""
Adjust the sharpness of the input image by a fixed or random degree. Degree of 0.0 gives a blurred image,
degree of 1.0 gives the original image, and degree of 2.0 gives a sharpened image.
Args:
degrees (Union[list, tuple], optional): Range of random sharpness adjustment degrees,
which must be non-negative. It should be in (min, max) format. If min=max, then
it is a single fixed magnitude operation. Default: ``(0.1, 1.9)``.
Raises:
TypeError : If `degrees` is not a list or a tuple.
ValueError: If `degrees` is negative.
ValueError: If `degrees` is in (max, min) format instead of (min, max).
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.RandomSharpness(degrees=(0.2, 1.9))]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomSharpness(degrees=(0, 0.6))(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_positive_degrees
def __init__(self, degrees=(0.1, 1.9)):
super().__init__()
self.degrees = degrees
def parse(self):
return cde.RandomSharpnessOperation(self.degrees)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be sharpness adjusted.
Returns:
PIL Image, sharpness adjusted image.
"""
return util.random_sharpness(img, self.degrees)
[docs]class RandomSolarize(ImageTensorOperation):
"""
Randomly selects a subrange within the specified threshold range and sets the pixel value within
the subrange to (255 - pixel).
Args:
threshold (tuple, optional): Range of random solarize threshold. Default: ``(0, 255)``.
Threshold values should always be in (min, max) format,
where min and max are integers in the range [0, 255], and min <= max. The pixel values
belonging to the [min, max] range will be inverted.
If min=max, then invert all pixel values greater than or equal min(max).
Raises:
TypeError : If `threshold` is not of type tuple.
ValueError: If `threshold` is not in range of [0, 255].
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.RandomSolarize(threshold=(10,100))]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RandomSolarize(threshold=(1, 10))(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_random_solarize
def __init__(self, threshold=(0, 255)):
super().__init__()
self.threshold = threshold
self.implementation = Implementation.C
def parse(self):
return cde.RandomSolarizeOperation(self.threshold)
[docs]class RandomVerticalFlip(ImageTensorOperation, PyTensorOperation):
"""
Randomly flip the input image vertically with a given probability.
Args:
prob (float, optional): Probability of the image being flipped, which
must be in range of [0.0, 1.0]. Default: ``0.5``.
Raises:
TypeError: If `prob` is not of type float.
ValueError: If `prob` is not in range [0.0, 1.0].
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.RandomVerticalFlip(0.25)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((2, 3))
>>> output = vision.RandomVerticalFlip(1.0)(data)
>>> print(output.shape, output.dtype)
(2, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_prob
def __init__(self, prob=0.5):
super().__init__()
self.prob = prob
def parse(self):
return cde.RandomVerticalFlipOperation(self.prob)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be vertically flipped.
Returns:
PIL Image, randomly vertically flipped image.
"""
return util.random_vertical_flip(img, self.prob)
[docs]class RandomVerticalFlipWithBBox(ImageTensorOperation):
"""
Flip the input image vertically, randomly with a given probability and adjust bounding boxes accordingly.
Args:
prob (float, optional): Probability of the image being flipped,
which must be in range of [0.0, 1.0]. Default: ``0.5``.
Raises:
TypeError: If `prob` is not of type float.
ValueError: If `prob` is not in range [0.0, 1.0].
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.float32)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> func = lambda img: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(np.float32))
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=[func],
... input_columns=["image"],
... output_columns=["image", "bbox"])
>>> transforms_list = [vision.RandomVerticalFlipWithBBox(0.20)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image", "bbox"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... print(item["bbox"].shape, item["bbox"].dtype)
... break
(100, 100, 3) float32
(1, 4) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.float32)
>>> func = lambda img: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(data.dtype))
>>> func_data, func_bboxes = func(data)
>>> output = vision.RandomVerticalFlipWithBBox(1)(func_data, func_bboxes)
>>> print(output[0].shape, output[0].dtype)
(100, 100, 3) float32
>>> print(output[1].shape, output[1].dtype)
(1, 4) float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_prob
def __init__(self, prob=0.5):
super().__init__()
self.prob = prob
self.implementation = Implementation.C
def parse(self):
return cde.RandomVerticalFlipWithBBoxOperation(self.prob)
[docs]class Rescale(ImageTensorOperation):
"""
Rescale the input image with the given rescale and shift. This operation will rescale the input image
with: output = image * rescale + shift.
Note:
This operation is executed on the CPU by default, but it is also supported
to be executed on the GPU or Ascend via heterogeneous acceleration.
Args:
rescale (float): Rescale factor.
shift (float): Shift factor.
Raises:
TypeError: If `rescale` is not of type float.
TypeError: If `shift` is not of type float.
Supported Platforms:
``CPU`` ``GPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.Rescale(1.0 / 255.0, -1.0)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.Rescale(1.0 / 255.0, -1.0)(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_rescale
def __init__(self, rescale, shift):
super().__init__()
self.rescale = rescale
self.shift = shift
self.implementation = Implementation.C
def parse(self):
return cde.RescaleOperation(self.rescale, self.shift)
[docs]class Resize(ImageTensorOperation, PyTensorOperation):
"""
Resize the input image to the given size with a given interpolation mode :class:`~.vision.Inter` .
Supports Ascend hardware acceleration and can be enabled through the `.device("Ascend")` method.
Args:
size (Union[int, Sequence[int]]): The output size of the resized image. The size value(s) must be positive.
If size is an integer, the smaller edge of the image will be resized to this value with
the same image aspect ratio.
If size is a sequence of length 2, it should be (height, width).
interpolation (Inter, optional): Image interpolation method defined by :class:`~.vision.Inter` .
Default: ``Inter.LINEAR``.
Raises:
TypeError: If `size` is not of type int or Sequence[int].
TypeError: If `interpolation` is not of type :class:`~.vision.Inter` .
ValueError: If `size` is not positive.
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> resize_op = vision.Resize([100, 75], Inter.BICUBIC)
>>> transforms_list = [resize_op]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 75, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.Resize([5, 5], Inter.BICUBIC)(data)
>>> print(output.shape, output.dtype)
(5, 5, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_resize_interpolation
def __init__(self, size, interpolation=Inter.LINEAR):
super().__init__()
self.py_size = size
if isinstance(size, int):
size = (size,)
self.c_size = size
self.interpolation = interpolation
if interpolation in [Inter.AREA, Inter.PILCUBIC]:
self.implementation = Implementation.C
elif interpolation == Inter.ANTIALIAS:
self.implementation = Implementation.PY
self.random = False
[docs] @check_device_target
def device(self, device_target="CPU"):
"""
Set the device for the current operator execution.
- When the device is Ascend, input/output shape should be limited from [4, 6] to [32768, 32768].
Args:
device_target (str, optional): The operator will be executed on this device. Currently supports
``CPU`` and ``Ascend`` . Default: ``CPU`` .
Raises:
TypeError: If `device_target` is not of type str.
ValueError: If `device_target` is not within the valid set of ['CPU', 'Ascend'].
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> resize_op = vision.Resize([100, 75], Inter.BICUBIC).device("Ascend")
>>> transforms_list = [resize_op]
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 75, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.Resize([25, 25], Inter.BICUBIC).device("Ascend")(data)
>>> print(output.shape, output.dtype)
(25, 25, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
self.device_target = device_target
if self.interpolation == Inter.ANTIALIAS and self.device_target == "Ascend":
raise ValueError("The current InterpolationMode is not supported by DVPP. It is {}."
.format(self.interpolation))
return self
def parse(self):
if self.interpolation == Inter.ANTIALIAS:
raise TypeError("The current InterpolationMode is not supported with NumPy input.")
return cde.ResizeOperation(self.c_size, Inter.to_c_type(self.interpolation), self.device_target)
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be resized.
Returns:
PIL Image, resized image.
"""
if self.interpolation in [Inter.AREA, Inter.PILCUBIC]:
raise TypeError("Current Interpolation is not supported with PIL input.")
return util.resize(img, self.py_size, Inter.to_python_type(self.interpolation))
[docs]class ResizedCrop(ImageTensorOperation):
"""
Crop the input image at a specific region and resize it to desired size.
Args:
top (int): Horizontal ordinate of the upper left corner of the crop region.
left (int): Vertical ordinate of the upper left corner of the crop region.
height (int): Height of the crop region.
width (int): Width of the cropp region.
size (Union[int, Sequence[int, int]]): The size of the output image.
If int is provided, the smaller edge of the image will be resized to this value,
keeping the image aspect ratio the same.
If Sequence[int, int] is provided, it should be (height, width).
interpolation (Inter, optional): Image interpolation method defined by :class:`~.vision.Inter` .
Default: ``Inter.BILINEAR``.
Raises:
TypeError: If `top` is not of type int.
ValueError: If `top` is negative.
TypeError: If `left` is not of type int.
ValueError: If `left` is negative.
TypeError: If `height` is not of type int.
ValueError: If `height` is not positive.
TypeError: If `width` is not of type int.
ValueError: If `width` is not positive.
TypeError: If `size` is not of type int or Sequence[int, int].
ValueError: If `size` is not posotive.
TypeError: If `interpolation` is not of type :class:`~.vision.Inter` .
RuntimeError: If shape of the input image is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> transforms_list = [vision.ResizedCrop(0, 0, 64, 64, (100, 75), Inter.BILINEAR)]
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 75, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.ResizedCrop(0, 0, 1, 1, (5, 5), Inter.BILINEAR)(data)
>>> print(output.shape, output.dtype)
(5, 5, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_resized_crop
def __init__(self, top, left, height, width, size, interpolation=Inter.BILINEAR):
super().__init__()
if isinstance(size, int):
size = (size, size)
self.top = top
self.left = left
self.height = height
self.width = width
self.size = size
self.interpolation = interpolation
self.implementation = Implementation.C
[docs] @check_device_target
def device(self, device_target="CPU"):
"""
Set the device for the current operator execution.
- When the device is Ascend, input/output shape should be limited from [4, 6] to [32768, 32768].
Args:
device_target (str, optional): The operator will be executed on this device. Currently supports
``CPU`` and ``Ascend`` . Default: ``CPU`` .
Raises:
TypeError: If `device_target` is not of type str.
ValueError: If `device_target` is not within the valid set of ['CPU', 'Ascend'].
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> resize_crop_op = vision.ResizedCrop(0, 0, 64, 64, (100, 75)).device("Ascend")
>>> transforms_list = [resize_crop_op]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 75, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.ResizedCrop(0, 0, 64, 64, (32, 16), Inter.BILINEAR).device("Ascend")(data)
>>> print(output.shape, output.dtype)
(32, 16, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
self.device_target = device_target
return self
def parse(self):
return cde.ResizedCropOperation(self.top, self.left, self.height,
self.width, self.size, Inter.to_c_type(self.interpolation), self.device_target)
[docs]class ResizeWithBBox(ImageTensorOperation):
"""
Resize the input image to the given size and adjust bounding boxes accordingly.
Args:
size (Union[int, Sequence[int]]): The output size of the resized image.
If size is an integer, smaller edge of the image will be resized to this value with
the same image aspect ratio.
If size is a sequence of length 2, it should be (height, width).
interpolation (Inter, optional): Image interpolation method defined by :class:`~.vision.Inter` .
Default: ``Inter.LINEAR``.
Raises:
TypeError: If `size` is not of type int or Sequence[int].
TypeError: If `interpolation` is not of type :class:`~.vision.Inter` .
ValueError: If `size` is not positive.
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.float32)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> func = lambda img: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(np.float32))
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=[func],
... input_columns=["image"],
... output_columns=["image", "bbox"])
>>> bbox_op = vision.ResizeWithBBox(50, Inter.NEAREST)
>>> transforms_list = [bbox_op]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image", "bbox"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... print(item["bbox"].shape, item["bbox"].dtype)
... break
(50, 50, 3) float32
(1, 4) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.float32)
>>> func = lambda img: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(data.dtype))
>>> func_data, func_bboxes = func(data)
>>> output = vision.ResizeWithBBox(100)(func_data, func_bboxes)
>>> print(output[0].shape, output[0].dtype)
(100, 100, 3) float32
>>> print(output[1].shape, output[1].dtype)
(1, 4) float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_resize_interpolation
def __init__(self, size, interpolation=Inter.LINEAR):
super().__init__()
self.size = size
self.interpolation = interpolation
self.implementation = Implementation.C
def parse(self):
size = self.size
if isinstance(size, int):
size = (size,)
return cde.ResizeWithBBoxOperation(size, Inter.to_c_type(self.interpolation))
[docs]class RgbToHsv(PyTensorOperation):
"""
Convert the input numpy.ndarray images from RGB to HSV.
Args:
is_hwc (bool): If ``True``, means the input image is in shape of <H, W, C> or <N, H, W, C>.
Otherwise, it is in shape of <C, H, W> or <N, C, H, W>. Default: ``False``.
Raises:
TypeError: If `is_hwc` is not of type bool.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.transforms import Compose
>>>
>>> # Use the transform in dataset pipeline mode
>>> transforms_list = Compose([vision.CenterCrop(20),
... vision.ToTensor(),
... vision.RgbToHsv()])
>>> # apply the transform to dataset through map function
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns="image")
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(3, 20, 20) float64
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.RgbToHsv(is_hwc=True)(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) float64
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_rgb_to_hsv
def __init__(self, is_hwc=False):
super().__init__()
self.is_hwc = is_hwc
self.random = False
self.implementation = Implementation.PY
def _execute_py(self, rgb_imgs):
"""
Execute method.
Args:
rgb_imgs (numpy.ndarray): RGB images to be converted.
Returns:
numpy.ndarray, converted HSV images.
"""
return util.rgb_to_hsvs(rgb_imgs, self.is_hwc)
[docs]class Rotate(ImageTensorOperation):
"""
Rotate the input image by specified degrees.
Args:
degrees (Union[int, float]): Rotation degrees.
resample (Inter, optional): Image interpolation method defined by :class:`~.vision.Inter` .
Default: ``Inter.NEAREST``.
expand (bool, optional): Optional expansion flag. Default: ``False``. If set to ``True``,
expand the output image to make it large enough to hold the entire rotated image.
If set to ``False`` or omitted, make the output image the same size as the input.
Note that the expand flag assumes rotation around the center and no translation.
center (tuple, optional): Optional center of rotation (a 2-tuple). Default: ``None``.
Origin is the top left corner. ``None`` sets to the center of the image.
fill_value (Union[int, tuple[int]], optional): Optional fill color for the area outside the rotated image.
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 `degrees` is not of type integer, float or sequence.
TypeError: If `resample` is not of type :class:`~.vision.Inter` .
TypeError: If `expand` is not of type bool.
TypeError: If `center` is not of type tuple.
TypeError: If `fill_value` is not of type int or tuple[int].
ValueError: If `fill_value` is not in range [0, 255].
RuntimeError: If given tensor shape is not <H, W> or <..., H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> transforms_list = [vision.Rotate(degrees=30.0, resample=Inter.NEAREST, expand=True)]
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(137, 137, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.Rotate(degrees=30.0, resample=Inter.NEAREST, expand=True)(data)
>>> print(output.shape, output.dtype)
(137, 137, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_rotate
def __init__(self, degrees, resample=Inter.NEAREST, expand=False, center=None, fill_value=0):
super().__init__()
if isinstance(degrees, (int, float)):
degrees = degrees % 360
if center is None:
center = ()
if isinstance(fill_value, int):
fill_value = tuple([fill_value] * 3)
self.degrees = degrees
self.resample = resample
self.expand = expand
self.center = center
self.fill_value = fill_value
self.implementation = Implementation.C
def parse(self):
return cde.RotateOperation(self.degrees, Inter.to_c_type(self.resample), self.expand, self.center,
self.fill_value)
[docs]class SlicePatches(ImageTensorOperation):
r"""
Slice Tensor to multiple patches in horizontal and vertical directions.
The usage scenario is suitable to large height and width Tensor. The Tensor
will keep the same if set both num_height and num_width to 1. And the
number of output tensors is equal to :math:`num\_height * num\_width`.
Args:
num_height (int, optional): The number of patches in vertical direction, which must be positive. Default: ``1``.
num_width (int, optional): The number of patches in horizontal direction, which must be positive.
Default: ``1``.
slice_mode (SliceMode, optional): A mode represents pad or drop. Default: ``SliceMode.PAD``.
It can be ``SliceMode.PAD``, ``SliceMode.DROP``.
fill_value (int, optional): The border width in number of pixels in
right and bottom direction if slice_mode is set to be SliceMode.PAD.
The `fill_value` must be in range [0, 255]. Default: ``0``.
Raises:
TypeError: If `num_height` is not of type integer.
TypeError: If `num_width` is not of type integer.
TypeError: If `slice_mode` is not of type Inter.
TypeError: If `fill_value` is not of type integer.
ValueError: If `num_height` is not positive.
ValueError: If `num_width` is not positive.
ValueError: If `fill_value` is not in range [0, 255].
RuntimeError: If given tensor shape is not <H, W> or <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> # default padding mode
>>> num_h, num_w = (1, 4)
>>> slice_patches_op = vision.SlicePatches(num_h, num_w)
>>> transforms_list = [slice_patches_op]
>>> cols = ['img' + str(x) for x in range(num_h*num_w)]
>>>
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list,
... input_columns=["image"],
... output_columns=cols)
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(len(item), item["img0"].shape, item["img0"].dtype)
... break
4 (100, 25, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.SlicePatches(1, 2)(data)
>>> print(np.array(output).shape, np.array(output).dtype)
(2, 100, 50, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_slice_patches
def __init__(self, num_height=1, num_width=1, slice_mode=SliceMode.PAD, fill_value=0):
super().__init__()
self.num_height = num_height
self.num_width = num_width
self.slice_mode = slice_mode
self.fill_value = fill_value
self.implementation = Implementation.C
def parse(self):
return cde.SlicePatchesOperation(self.num_height, self.num_width,
SliceMode.to_c_type(self.slice_mode), self.fill_value)
[docs]class Solarize(ImageTensorOperation):
"""
Solarize the image by inverting all pixel values within the threshold.
Args:
threshold (Union[float, Sequence[float, float]]): Range of solarize threshold, should always
be in (min, max) format, where min and max are integers in range of [0, 255], and min <= max.
The pixel values belonging to the [min, max] range will be inverted.
If a single value is provided or min=max, then invert all pixel values greater than or equal min(max).
Raises:
TypeError: If `threshold` is not of type float or Sequence[float, float].
ValueError: If `threshold` is not in range of [0, 255].
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.Solarize(threshold=(10, 100))]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.Solarize(threshold=(1, 10))(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_solarize
def __init__(self, threshold):
super().__init__()
if isinstance(threshold, (float, int)):
threshold = (threshold, threshold)
self.threshold = threshold
self.implementation = Implementation.C
def parse(self):
return cde.SolarizeOperation(self.threshold)
[docs]class TenCrop(PyTensorOperation):
"""
Crop the given image into one central crop and four corners with the flipped version of these.
Args:
size (Union[int, Sequence[int, int]]): The size of the cropped image.
If a single integer is provided, a square of size (size, size) will be cropped with this value.
If a sequence of length 2 is provided, an image of size (height, width) will be cropped.
use_vertical_flip (bool, optional): If ``True``, flip the images vertically. Otherwise, flip them
horizontally. Default: ``False``.
Raises:
TypeError: If `size` is not of type integer or sequence of integer.
TypeError: If `use_vertical_flip` is not of type boolean.
ValueError: If `size` is not positive.
Supported Platforms:
``CPU``
Examples:
>>> import os
>>> import numpy as np
>>> from PIL import Image, ImageDraw
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.transforms import Compose
>>>
>>> # Use the transform in dataset pipeline mode
>>> class MyDataset:
... def __init__(self):
... self.data = []
... img = Image.new("RGB", (300, 300), (255, 255, 255))
... draw = ImageDraw.Draw(img)
... draw.ellipse(((0, 0), (100, 100)), fill=(255, 0, 0), outline=(255, 0, 0), width=5)
... img.save("./1.jpg")
... data = np.fromfile("./1.jpg", np.uint8)
... self.data.append(data)
...
... def __getitem__(self, index):
... return self.data[0]
...
... def __len__(self):
... return 5
>>>
>>> my_dataset = MyDataset()
>>> generator_dataset = ds.GeneratorDataset(my_dataset, column_names="image")
>>> transforms_list = Compose([vision.Decode(to_pil=True),
... vision.TenCrop(size=200),
... # 4D stack of 10 images
... lambda *images: np.stack([vision.ToTensor()(image) for image in images])])
>>> # apply the transform to dataset through map function
>>> generator_dataset = generator_dataset.map(operations=transforms_list, input_columns="image")
>>> for item in generator_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(10, 3, 200, 200) float32
>>> os.remove("./1.jpg")
>>>
>>> # Use the transform in eager mode
>>> img = Image.new("RGB", (300, 300), (255, 255, 255))
>>> draw = ImageDraw.Draw(img)
>>> draw.polygon([(50, 50), (150, 50), (100, 150)], fill=(0, 255, 0), outline=(0, 255, 0))
>>> img.save("./2.jpg")
>>> data = Image.open("./2.jpg")
>>> output = vision.TenCrop(size=200)(data)
>>> print(len(output), np.array(output[0]).shape, np.array(output[0]).dtype)
10 (200, 200, 3) uint8
>>> os.remove("./2.jpg")
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_ten_crop
def __init__(self, size, use_vertical_flip=False):
super().__init__()
if isinstance(size, int):
size = (size, size)
self.size = size
self.use_vertical_flip = use_vertical_flip
self.random = False
self.implementation = Implementation.PY
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be cropped.
Returns:
tuple, a tuple of 10 PIL Image, in order of top_left, top_right, bottom_left, bottom_right, center
of the original image and top_left, top_right, bottom_left, bottom_right, center of the flipped image.
"""
return util.ten_crop(img, self.size, self.use_vertical_flip)
[docs]class ToNumpy(PyTensorOperation):
"""
Convert the PIL input image to numpy.ndarray image.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.transforms import Compose
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> # Use ToNumpy to explicitly select C++ implementation of subsequent op
>>> transforms_list = Compose([vision.RandomHorizontalFlip(0.5),
... vision.ToNumpy(),
... vision.Resize((50, 60))])
>>> # apply the transform to dataset through map function
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns="image")
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(50, 60, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = list(np.random.randint(0, 255, size=(32, 32, 3, 3)).astype(np.int32))
>>> output = vision.ToNumpy()(data)
>>> print(type(output), output.shape, output.dtype)
<class 'numpy.ndarray'> (32, 32, 3, 3) int32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
def __init__(self):
super().__init__()
self.random = False
# Use "Implementation.C" to indicate to select C++ implementation for next op in transforms list
self.implementation = Implementation.C
def _execute_py(self, img):
"""
Execute method.
Args:
img (PIL Image): Image to be converted to numpy.ndarray.
Returns:
Image converted to numpy.ndarray
"""
return np.array(img)
[docs]class ToPIL(PyTensorOperation):
"""
Convert the input decoded numpy.ndarray image to PIL Image.
Raises:
TypeError: If the input image is not of type :class:`numpy.ndarray` or `PIL.Image.Image` .
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.transforms import Compose
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> # data is already decoded, but not in PIL Image format
>>> transforms_list = Compose([vision.ToPIL(),
... vision.RandomHorizontalFlip(0.5),
... vision.ToTensor()])
>>> # apply the transform to dataset through map function
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns="image")
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(3, 100, 100) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.ToPIL()(data)
>>> print(type(output), np.array(output).shape, np.array(output).dtype)
<class 'PIL.Image.Image'> (100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
def __init__(self):
super().__init__()
self.random = False
self.implementation = Implementation.PY
def _execute_py(self, img):
"""
Execute method.
Args:
img (numpy.ndarray): Decoded numpy.ndarray image to be converted to PIL Image.
Returns:
PIL Image, converted PIL Image.
"""
return util.to_pil(img)
[docs]class ToTensor(ImageTensorOperation):
"""
Convert the input PIL Image or numpy.ndarray to numpy.ndarray of the desired dtype, rescale the pixel value
range from [0, 255] to [0.0, 1.0] and change the shape from <H, W, C> to <C, H, W>.
Args:
output_type (Union[mindspore.dtype, numpy.dtype], optional): The desired dtype of the output image.
Default: ``np.float32`` .
Raises:
TypeError: If the input image is not of type `PIL.Image.Image` or :class:`numpy.ndarray` .
TypeError: If dimension of the input image is not 2 or 3.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.transforms import Compose
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> # create a list of transformations to be applied to the "image" column of each data row
>>> transforms_list = Compose([vision.RandomHorizontalFlip(0.5),
... vision.ToTensor()])
>>> # apply the transform to dataset through map function
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns="image")
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(3, 100, 100) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.ToTensor()(data)
>>> print(output.shape, output.dtype)
(3, 100, 100) float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_to_tensor
def __init__(self, output_type=np.float32):
super().__init__()
if isinstance(output_type, typing.Type):
output_type = mstype_to_detype(output_type)
else:
output_type = nptype_to_detype(output_type)
self.output_type = str(output_type)
self.random = False
self.implementation = Implementation.C
def parse(self):
return cde.ToTensorOperation(self.output_type)
[docs]class ToType(TypeCast):
"""
Cast the input to a given MindSpore data type or NumPy data type.
It is the same as that of :class:`mindspore.dataset.transforms.TypeCast` .
Note:
This operation is executed on the CPU by default, but it is also supported
to be executed on the GPU or Ascend via heterogeneous acceleration.
Args:
data_type (Union[mindspore.dtype, numpy.dtype]): The desired data type of the output image,
such as ``numpy.float32`` .
Raises:
TypeError: If `data_type` is not of type :class:`mindspore.dtype` or :class:`numpy.dtype` .
Supported Platforms:
``CPU`` ``GPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> import numpy as np
>>> from mindspore.dataset.transforms import Compose
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = Compose([vision.RandomHorizontalFlip(0.5),
... vision.ToTensor(),
... vision.ToType(np.float32)])
>>> # apply the transform to dataset through map function
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns="image")
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(3, 100, 100) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.array([2.71606445312564e-03, 6.3476562564e-03]).astype(np.float64)
>>> output = vision.ToType(np.float32)(data)
>>> print(output, output.dtype)
[0.00271606 0.00634766] float32
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
[docs]class TrivialAugmentWide(ImageTensorOperation):
"""
Apply TrivialAugmentWide data augmentation method on the input image.
Refer to
`TrivialAugmentWide: Tuning-free Yet State-of-the-Art Data Augmentation <https://arxiv.org/abs/2103.10158>`_ .
Only support 3-channel RGB image.
Args:
num_magnitude_bins (int, optional): The number of different magnitude values,
must be greater than or equal to 2. Default: ``31``.
interpolation (Inter, optional): Image interpolation method defined by :class:`~.vision.Inter` .
Default: ``Inter.NEAREST``.
fill_value (Union[int, tuple[int, int, int]], optional): Pixel fill value for the area outside the
transformed image, must be in range of [0, 255]. Default: ``0``.
If int is provided, pad all RGB channels with this value.
If tuple[int, int, int] is provided, pad R, G, B channels respectively.
Raises:
TypeError: If `num_magnitude_bins` is not of type int.
ValueError: If `num_magnitude_bins` is less than 2.
TypeError: If `interpolation` not of type :class:`~.vision.Inter` .
TypeError: If `fill_value` is not of type int or tuple[int, int, int].
ValueError: If `fill_value` is not in range of [0, 255].
RuntimeError: If shape of the input image is not <H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>> from mindspore.dataset.vision import Inter
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.TrivialAugmentWide(num_magnitude_bins=31,
... interpolation=Inter.NEAREST,
... fill_value=0)]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.TrivialAugmentWide()(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
@check_trivial_augment_wide
def __init__(self, num_magnitude_bins=31, interpolation=Inter.NEAREST, fill_value=0):
super().__init__()
self.num_magnitude_bins = num_magnitude_bins
self.interpolation = interpolation
if isinstance(fill_value, int):
fill_value = tuple([fill_value] * 3)
self.fill_value = fill_value
self.implementation = Implementation.C
def parse(self):
return cde.TrivialAugmentWideOperation(self.num_magnitude_bins, Inter.to_c_type(self.interpolation),
self.fill_value)
[docs]class VerticalFlip(ImageTensorOperation):
"""
Flip the input image vertically.
Raises:
RuntimeError: If given tensor shape is not <H, W> or <..., H, W, C>.
Supported Platforms:
``CPU``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> transforms_list = [vision.VerticalFlip()]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.VerticalFlip()(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
def __init__(self):
super().__init__()
self.implementation = Implementation.C
[docs] @check_device_target
def device(self, device_target="CPU"):
"""
Set the device for the current operator execution.
- When the device is Ascend, input shape should be limited from [4, 6] to [8192, 4096].
Args:
device_target (str, optional): The operator will be executed on this device. Currently supports
``CPU`` and ``Ascend`` . Default: ``CPU`` .
Raises:
TypeError: If `device_target` is not of type str.
ValueError: If `device_target` is not within the valid set of ['CPU', 'Ascend'].
Supported Platforms:
``CPU`` ``Ascend``
Examples:
>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(1, 100, 100, 3)).astype(np.uint8)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> vertical_flip_op = vision.VerticalFlip().device("Ascend")
>>> transforms_list = [vertical_flip_op]
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=transforms_list, input_columns=["image"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
... print(item["image"].shape, item["image"].dtype)
... break
(100, 100, 3) uint8
>>>
>>> # Use the transform in eager mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.uint8)
>>> output = vision.VerticalFlip().device("Ascend")(data)
>>> print(output.shape, output.dtype)
(100, 100, 3) uint8
Tutorial Examples:
- `Illustration of vision transforms
<https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/samples/dataset/vision_gallery.html>`_
"""
self.device_target = device_target
return self
def parse(self):
return cde.VerticalFlipOperation(self.device_target)
def not_random(func):
"""
Specify the function as "not random", i.e., it produces deterministic result.
A Python function can only be cached after it is specified as "not random".
"""
func.random = False
return func