Augmentation

Functions used to increase and diversify a dataset by duplicating and transforming images. Available transformations are:

• Identity

• Transpose

• Inverse transpose

• Horizontal flip

• Vertical flip

• 90° rotation

• 180° rotation

• 270° rotation

Apply a random transformation on a 2D image:

• bigfish.stack.augment_2d()

• bigfish.stack.augment_2d_function()

Apply all the possible transformations on a 2D image:

• bigfish.stack.augment_8_times()

• bigfish.stack.augment_8_times_reversed()

bigfish.stack.augment_2d(image)

Augment an image applying a random operation.

Parameters:

imagenp.ndarray

Image to augment with shape (y, x, channels) or (y, x, channels).

Returns:

image_augmentednp.ndarray

Image augmented with shape (y, x, channels).

bigfish.stack.augment_2d_function(identity=False)

Choose a random operation to augment a 2-d image.

Parameters:

identitybool

Return identity function instead of a random transformation.

Returns:

random_operationcallable

Function to transform a 2-d image.

bigfish.stack.augment_8_times(image)

Apply every transformation to a 2-d image.

Parameters:

imagenp.ndarray

Image to augment with shape (y, x, channels).

Returns:

images_augmentedList[np.ndarray]

List of images augmented with shape (y, x, channels).

bigfish.stack.augment_8_times_reversed(images_augmented)

Apply every transformation back to return the original 2-d image.

Parameters:

images_augmentedList[np.ndarray]

List of images augmented with shape (y, x, channels).

Returns:

images_originalList[np.ndarray]

List of original images with shape (y, x, channels).