p47068915
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

!1463 Updated UA, RandSharp and RandColor parameter check, Updated UA code and description. 

Merge pull request !1463 from alashkari/ua-ops-v2
This commit is contained in:
mindspore-ci-bot 2020-05-27 15:16:56 +08:00 committed by Gitee
parent c0e6da7adb 2d1e1f4053
commit 2e2c516423
4 changed files with 72 additions and 41 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
mindspore/dataset/transforms/vision/c_transforms.py
View File

@ -45,7 +45,7 @@ import mindspore._c_dataengine as cde
from .utils import Inter, Border
from .validators import check_prob, check_crop, check_resize_interpolation, check_random_resize_crop, \
check_normalize_c, check_random_crop, check_random_color_adjust, check_random_rotation, \
check_resize, check_rescale, check_pad, check_cutout, check_uniform_augmentation
check_resize, check_rescale, check_pad, check_cutout, check_uniform_augment_cpp
DE_C_INTER_MODE = {Inter.NEAREST: cde.InterpolationMode.DE_INTER_NEAREST_NEIGHBOUR,
Inter.LINEAR: cde.InterpolationMode.DE_INTER_LINEAR,
@ -456,7 +456,7 @@ class UniformAugment(cde.UniformAugOp):
Args:
operations: list of C++ operations (python OPs are not accepted).
num_ops (int): number of OPs to be selected and applied.
num_ops (int, optional): number of OPs to be selected and applied (default=2).
Examples:
>>> transforms_list = [c_transforms.RandomHorizontalFlip(),
@ -470,7 +470,7 @@ class UniformAugment(cde.UniformAugOp):
>>> operations=transforms_all, num_parallel_workers=1)
"""
@check_uniform_augmentation
@check_uniform_augment_cpp
def __init__(self, operations, num_ops=2):
self.operations = operations
self.num_ops = num_ops

6
mindspore/dataset/transforms/vision/py_transforms.py
View File

@ -32,7 +32,7 @@ from .validators import check_prob, check_crop, check_resize_interpolation, chec
check_normalize_py, check_random_crop, check_random_color_adjust, check_random_rotation, \
check_transforms_list, check_random_apply, check_ten_crop, check_num_channels, check_pad, \
check_random_perspective, check_random_erasing, check_cutout, check_linear_transform, check_random_affine, \
check_mix_up
check_mix_up, check_positive_degrees, check_uniform_augment_py
from .utils import Inter, Border
DE_PY_INTER_MODE = {Inter.NEAREST: Image.NEAREST,
@ -1329,6 +1329,7 @@ class RandomColor:
>>> py_transforms.ToTensor()])
"""
@check_positive_degrees
def __init__(self, degrees=(0.1, 1.9)):
self.degrees = degrees
@ -1361,6 +1362,7 @@ class RandomSharpness:
"""
@check_positive_degrees
def __init__(self, degrees=(0.1, 1.9)):
self.degrees = degrees
@ -1458,6 +1460,7 @@ class UniformAugment:
Uniformly select and apply a number of transforms sequentially from
a list of transforms. Randomly assigns a probability to each transform for
each image to decide whether apply it or not.
All the transforms in transform list must have the same input/output data type.
Args:
transforms (list): List of transformations to be chosen from to apply.
@ -1473,6 +1476,7 @@ class UniformAugment:
>>> py_transforms.ToTensor()])
"""
@check_uniform_augment_py
def __init__(self, transforms, num_ops=2):
self.transforms = transforms
self.num_ops = num_ops

40
mindspore/dataset/transforms/vision/py_transforms_util.py
View File

@ -1427,17 +1427,6 @@ def random_color(img, degrees):
if not is_pil(img):
raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
if isinstance(degrees, (list, tuple)):
if len(degrees) != 2:
raise ValueError("Degrees must be a sequence length 2.")
if degrees[0] < 0:
raise ValueError("Degree value must be non-negative.")
if degrees[0] > degrees[1]:
raise ValueError("Degrees should be in (min,max) format. Got (max,min).")
else:
raise TypeError("Degrees must be a sequence in (min,max) format.")
v = (degrees[1] - degrees[0]) * random.random() + degrees[0]
return ImageEnhance.Color(img).enhance(v)
@ -1459,17 +1448,6 @@ def random_sharpness(img, degrees):
if not is_pil(img):
raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
if isinstance(degrees, (list, tuple)):
if len(degrees) != 2:
raise ValueError("Degrees must be a sequence length 2.")
if degrees[0] < 0:
raise ValueError("Degree value must be non-negative.")
if degrees[0] > degrees[1]:
raise ValueError("Degrees should be in (min,max) format. Got (max,min).")
else:
raise TypeError("Degrees must be a sequence in (min,max) format.")
v = (degrees[1] - degrees[0]) * random.random() + degrees[0]
return ImageEnhance.Sharpness(img).enhance(v)
@ -1537,6 +1515,7 @@ def uniform_augment(img, transforms, num_ops):
Uniformly select and apply a number of transforms sequentially from
a list of transforms. Randomly assigns a probability to each transform for
each image to decide whether apply it or not.
All the transforms in transform list must have the same input/output data type.
Args:
img: Image to be applied transformation.
@ -1545,23 +1524,14 @@ def uniform_augment(img, transforms, num_ops):
Returns:
img, Transformed image.
"""
if transforms is None:
raise ValueError("transforms is not provided.")
if not isinstance(transforms, list):
raise ValueError("The transforms needs to be a list.")
if not isinstance(num_ops, int):
raise ValueError("Number of operations should be a positive integer.")
if num_ops < 1:
raise ValueError("Number of operators should equal or greater than one.")
for _ in range(num_ops):
AugmentOp = random.choice(transforms)
op_idx = np.random.choice(len(transforms), size=num_ops, replace=False)
for idx in op_idx:
AugmentOp = transforms[idx]
pr = random.random()
if random.random() < pr:
img = AugmentOp(img.copy())
transforms.remove(AugmentOp)
return img

61
mindspore/dataset/transforms/vision/validators.py
View File

@ -816,8 +816,8 @@ def check_rescale(method):
return new_method
def check_uniform_augmentation(method):
"""Wrapper method to check the parameters of UniformAugmentation."""
def check_uniform_augment_cpp(method):
"""Wrapper method to check the parameters of UniformAugment cpp op."""
@wraps(method)
def new_method(self, *args, **kwargs):
@ -850,3 +850,60 @@ def check_uniform_augmentation(method):
return method(self, **kwargs)
return new_method
def check_uniform_augment_py(method):
"""Wrapper method to check the parameters of python UniformAugment op."""
@wraps(method)
def new_method(self, *args, **kwargs):
transforms, num_ops = (list(args) + 2 * [None])[:2]
if "transforms" in kwargs:
transforms = kwargs.get("transforms")
if transforms is None:
raise ValueError("transforms is not provided.")
if not transforms:
raise ValueError("transforms list is empty.")
check_list(transforms)
for transform in transforms:
if isinstance(transform, TensorOp):
raise ValueError("transform list only accepts Python operations.")
kwargs["transforms"] = transforms
if "num_ops" in kwargs:
num_ops = kwargs.get("num_ops")
if num_ops is not None:
check_type(num_ops, int)
check_positive(num_ops)
if num_ops > len(transforms):
raise ValueError("num_ops cannot be greater than the length of transforms list.")
kwargs["num_ops"] = num_ops
return method(self, **kwargs)
return new_method
def check_positive_degrees(method):
"""A wrapper method to check degrees parameter in RandSharpness and RandColor"""
@wraps(method)
def new_method(self, *args, **kwargs):
degrees = (list(args) + [None])[0]
if "degrees" in kwargs:
degrees = kwargs.get("degrees")
if degrees is not None:
if isinstance(degrees, (list, tuple)):
if len(degrees) != 2:
raise ValueError("Degrees must be a sequence with length 2.")
if degrees[0] < 0:
raise ValueError("Degrees range must be non-negative.")
if degrees[0] > degrees[1]:
raise ValueError("Degrees should be in (min,max) format. Got (max,min).")
else:
raise TypeError("Degrees must be a sequence in (min,max) format.")
return method(self, **kwargs)
return new_method