utilities.segmentation_utils.ImagePreprocessor module

class PreprocessingQueue(queue: list[Callable], arguments: list[Dict])

Bases: object

object to initialize a preprocessing queue

Parameters

queue list:

list of functions to be applied

arguments list[dict]:

list of arguments to be passed to the functions

arguments: list[Dict]

get_queue_length() → int

Returns the length of the queue

Returns

return int:

length of the queue

queue: list[Callable]

update_seed(seed)

Changes the seed of the queue

Parameters

seed int:

seed to be changed to

class PreprocessorInterface(*args, **kwargs)

Bases: Protocol

arguments: list[Dict]

get_queue_length() → int

queue: list[Callable]

update_seed(seed: int) → None

augmentation_pipeline(image, mask, input_size: tuple[int, int], output_size: tuple[int, int], image_queue: PreprocessingQueue, mask_queue: PreprocessingQueue, output_reshape: tuple[int, int] | None = None, channels: int = 3, seed: int = 0) → tuple[tensorflow.python.framework.ops.Tensor, tensorflow.python.framework.ops.Tensor]

Function that can execute a set of predifined augmentation functions stored in a PreprocessingQueue object. It augments both the image and the mask with the same functions and arguments.

Parameters

tf.Tensor image:

The image to be processed

tf.Tensor mask:

The mask to be processed

tuple(int, int) input_size:

Input size of the image

tuple(int, int) output_size:

Output size of the image

Keyword Arguments

tuple(int, int), optional output_reshape:

In case the image is a column vector, this is the shape it should be reshaped to. Defaults to None.

PreprocessingQueue, optional mask_queue image_queue:

Augmentation processing queue for images, defaults to None

PreprocessingQueue, optional mask_queue:

Augmentation processing queue for masks, defaults to None

int, optional channels:

Number of bands in the image, defaults to 3 :int, optional seed: The seed to be used in the pipeline, defaults to 0

Raises

raises ValueError:

If only one of the queues is passed

Returns

return tuple(tf.Tensor, tf.Tensor):

tuple of the processed image and mask

flatten(image, input_size, channels=1) → Tensor

Function that flattens an image preserving the number of channels

Parameters

tf.Tensor image:

image to be flattened

tuple(int, int) input_size:

input size of the image

Keyword Arguments

int, optional channels:

number of chanels to preserve, defaults to 1

Returns

return tf.Tensor:

flattened image

generate_default_queue(seed=0) → tuple[utilities.segmentation_utils.ImagePreprocessor.PreprocessingQueue, utilities.segmentation_utils.ImagePreprocessor.PreprocessingQueue]

Generates the default processing queue

Keyword Arguments

seed int:

seed to be used for the random functions

Returns

return PreprocessingQueue:

default queue

onehot_encode(masks, output_size, num_classes) → Tensor

Function that one-hot encodes masks

Batch(tf.Tensor) masks:

Masks to be encoded

Tuple(int, int) output_size:

Output size of the masks

Int num_classes:

Number of classes in the masks

Returns

return tf.Tensor:

Batch of one-hot encoded masks

random_flip_left_right(image, seed=0) → Tensor

Function that randomly flips an image left or right

Parameters

tf.Tensor image:

image to be flipped

Returns

return tf.Tensor:

flipped image

random_flip_up_down(image, seed=0) → Tensor

Function that randomly flips an image up or down

Parameters

tf.Tensor image:

image to be flipped

Returns

return tf.Tensor:

flipped image