dgbpy.transforms

Module Contents

Classes

BaseTransform	Base class for all transforms. All transforms should inherit from this class.
Flip	Base class for all transforms. All transforms should inherit from this class.
GaussianNoise	Base class for all transforms. All transforms should inherit from this class.
Rotate	Base class for all transforms. All transforms should inherit from this class.
Translate	Base class for all transforms. All transforms should inherit from this class.
FlipPolarity	Base class for all transforms. All transforms should inherit from this class.
ScaleTransform	Base class for all transforms. All transforms should inherit from this class.
Normalization	Base class for all transforms. All transforms should inherit from this class.
StandardScaler	Base class for all transforms. All transforms should inherit from this class.
MinMaxScaler	Base class for all transforms. All transforms should inherit from this class.
TransformCompose	Applies all the transform from a list to the data.
TransformMultiplier	Helps understand how many times each transform would multiply the data.

Functions

hasOpenCV()

Attributes

scale_transforms
all_transforms

class dgbpy.transforms.BaseTransform(p=0.2)

Bases: abc.ABC

Base class for all transforms. All transforms should inherit from this class. To create a new transform, inherit from this class and implement the transform_label and transform functions.

Example:

class MyTransform(BaseTransform):

def __init__(self, p=0.2):

super().__init__() self.p = p self.multiplier = 1

def transform_label(self, info):

return dgbhdf5.isImg2Img(info)

def transform(self, arr):

return arr + 1

Note:

self.multiplier is used to determine the number of times the transform should be applied. For example, if self.multiplier = 2, then the transform will be applied twice

for cases like Flip that can be done in multiple directions.

can_apply(self, info)

Returns True if the transform can be applied to the data.

abstract transform_label(self, info)

Returns True if the transform should be applied to the label

abstract transform(self, arr)

Returns the transformed array. This should hold the logic for the transform.

__call__(self, image=None, label=None, **kwargs)

This is the main function that is called when the transform is applied.

class dgbpy.transforms.Flip(p=0.2)

Bases: BaseTransform

Base class for all transforms. All transforms should inherit from this class. To create a new transform, inherit from this class and implement the transform_label and transform functions.

Example:

class MyTransform(BaseTransform):

def __init__(self, p=0.2):

super().__init__() self.p = p self.multiplier = 1

def transform_label(self, info):

return dgbhdf5.isImg2Img(info)

def transform(self, arr):

return arr + 1

Note:

self.multiplier is used to determine the number of times the transform should be applied. For example, if self.multiplier = 2, then the transform will be applied twice

for cases like Flip that can be done in multiple directions.

set_multiplier(self, info)

transform_label(self, info)

Returns True if the transform should be applied to the label

transform_pars(self, inp_shape)

transform(self, arr)

Returns the transformed array. This should hold the logic for the transform.

class dgbpy.transforms.GaussianNoise(p=0.2, std=0.1)

Bases: BaseTransform

Base class for all transforms. All transforms should inherit from this class. To create a new transform, inherit from this class and implement the transform_label and transform functions.

Example:

class MyTransform(BaseTransform):

def __init__(self, p=0.2):

super().__init__() self.p = p self.multiplier = 1

def transform_label(self, info):

return dgbhdf5.isImg2Img(info)

def transform(self, arr):

return arr + 1

Note:

self.multiplier is used to determine the number of times the transform should be applied. For example, if self.multiplier = 2, then the transform will be applied twice

for cases like Flip that can be done in multiple directions.

transform_label(self, info)

Returns True if the transform should be applied to the label

transform(self, arr)

Returns the transformed array. This should hold the logic for the transform.

dgbpy.transforms.hasOpenCV()

class dgbpy.transforms.Rotate(p=0.2, angle=15)

Bases: BaseTransform

Base class for all transforms. All transforms should inherit from this class. To create a new transform, inherit from this class and implement the transform_label and transform functions.

Example:

class MyTransform(BaseTransform):

def __init__(self, p=0.2):

super().__init__() self.p = p self.multiplier = 1

def transform_label(self, info):

return dgbhdf5.isImg2Img(info)

def transform(self, arr):

return arr + 1

Note:

self.multiplier is used to determine the number of times the transform should be applied. For example, if self.multiplier = 2, then the transform will be applied twice

for cases like Flip that can be done in multiple directions.

transform_label(self, info)

Returns True if the transform should be applied to the label

transform(self, arr)

Returns the transformed array. This should hold the logic for the transform.

transform_2d(self, arr, angle)

transform_3d(self, arr, angle)

class dgbpy.transforms.Translate(p=0.15, percent=20)

Bases: BaseTransform

Base class for all transforms. All transforms should inherit from this class. To create a new transform, inherit from this class and implement the transform_label and transform functions.

Example:

class MyTransform(BaseTransform):

def __init__(self, p=0.2):

super().__init__() self.p = p self.multiplier = 1

def transform_label(self, info):

return dgbhdf5.isImg2Img(info)

def transform(self, arr):

return arr + 1

Note:

self.multiplier is used to determine the number of times the transform should be applied. For example, if self.multiplier = 2, then the transform will be applied twice

for cases like Flip that can be done in multiple directions.

transform_label(self, info)

Returns True if the transform should be applied to the label

transform(self, arr)

Returns the transformed array. This should hold the logic for the transform.

class dgbpy.transforms.FlipPolarity(p=0.2)

Bases: BaseTransform

Base class for all transforms. All transforms should inherit from this class. To create a new transform, inherit from this class and implement the transform_label and transform functions.

Example:

class MyTransform(BaseTransform):

def __init__(self, p=0.2):

super().__init__() self.p = p self.multiplier = 1

def transform_label(self, info):

return dgbhdf5.isImg2Img(info)

def transform(self, arr):

return arr + 1

Note:

self.multiplier is used to determine the number of times the transform should be applied. For example, if self.multiplier = 2, then the transform will be applied twice

for cases like Flip that can be done in multiple directions.

transform_label(self, info)

Check if the label should be transformed

transform(self, arr)

FlipPolarity transform logic

class dgbpy.transforms.ScaleTransform

Bases: BaseTransform

Base class for all transforms. All transforms should inherit from this class. To create a new transform, inherit from this class and implement the transform_label and transform functions.

Example:

class MyTransform(BaseTransform):

def __init__(self, p=0.2):

super().__init__() self.p = p self.multiplier = 1

def transform_label(self, info):

return dgbhdf5.isImg2Img(info)

def transform(self, arr):

return arr + 1

Note:

self.multiplier is used to determine the number of times the transform should be applied. For example, if self.multiplier = 2, then the transform will be applied twice

for cases like Flip that can be done in multiple directions.

transform_label(self, info)

Returns True if the transform should be applied to the label

class dgbpy.transforms.Normalization

Bases: ScaleTransform, BaseTransform

Base class for all transforms. All transforms should inherit from this class. To create a new transform, inherit from this class and implement the transform_label and transform functions.

Example:

class MyTransform(BaseTransform):

def __init__(self, p=0.2):

super().__init__() self.p = p self.multiplier = 1

def transform_label(self, info):

return dgbhdf5.isImg2Img(info)

def transform(self, arr):

return arr + 1

Note:

self.multiplier is used to determine the number of times the transform should be applied. For example, if self.multiplier = 2, then the transform will be applied twice

for cases like Flip that can be done in multiple directions.

transform(self, arr)

Returns the transformed array. This should hold the logic for the transform.

class dgbpy.transforms.StandardScaler

Bases: ScaleTransform, BaseTransform

Base class for all transforms. All transforms should inherit from this class. To create a new transform, inherit from this class and implement the transform_label and transform functions.

Example:

class MyTransform(BaseTransform):

def __init__(self, p=0.2):

super().__init__() self.p = p self.multiplier = 1

def transform_label(self, info):

return dgbhdf5.isImg2Img(info)

def transform(self, arr):

return arr + 1

Note:

self.multiplier is used to determine the number of times the transform should be applied. For example, if self.multiplier = 2, then the transform will be applied twice

for cases like Flip that can be done in multiple directions.

transform(self, arr)

Returns the transformed array. This should hold the logic for the transform.

class dgbpy.transforms.MinMaxScaler

Bases: ScaleTransform, BaseTransform

Base class for all transforms. All transforms should inherit from this class. To create a new transform, inherit from this class and implement the transform_label and transform functions.

Example:

class MyTransform(BaseTransform):

def __init__(self, p=0.2):

super().__init__() self.p = p self.multiplier = 1

def transform_label(self, info):

return dgbhdf5.isImg2Img(info)

def transform(self, arr):

return arr + 1

Note:

self.multiplier is used to determine the number of times the transform should be applied. For example, if self.multiplier = 2, then the transform will be applied twice

for cases like Flip that can be done in multiple directions.

transform(self, arr)

Returns the transformed array. This should hold the logic for the transform.

dgbpy.transforms.scale_transforms

dgbpy.transforms.all_transforms

class dgbpy.transforms.TransformCompose(transforms, info, ndims, create_copy=False)

Applies all the transform from a list to the data.

set_params(self)

Helps understand how many times each transform would multiply the data.

set_uniform_generator_seed(self, seed, nsamples)

Sets the seed sample to receive the same type of tranform for each training.

passModuleCheck(self, transform_i)

Checks if the transform can be applied.

_readTransforms(self, transforms)

Read and initialize the transforms and checks if they can be applied.

set_multiplier(self, transforms)

transformLabel(self)

Checks if the transform can be applied to the label.

copy_config(self, transform_idx)

Configures the copy method.

use_copy_method(self, copy_prob)

Checks if the copy method is being used.

__call__(self, image, label, prob_idx, transform_idx=None)

Applies all the transforms to the data.

Args:

image: sample label: label prob_idx: index of the current sample used to choose the uniform probability when using seed transform_idx: value to be used for mixed transforms

class dgbpy.transforms.TransformMultiplier

Helps understand how many times each transform would multiply the data.

add(self, multiplier)

Adds the multiplier to the process map.

__len__(self)