Aug image

ImageAugmentation

The Image Augmentation class performs diverse augmentation methods on given numpy array. The class acts as an easy to use function/interface for applying all types of augmentations with just one function call.

The class can be configured beforehand by selecting desired augmentation techniques and method ranges or strength. Afterwards, the class is passed to the DataGenerator which utilizes it during batch generation.

The specific configurations of selected methods can be adjusted by class variables.

Build on top of the library

Albumentations - https://github.com/albumentations-team/albumentations

Source code in aucmedi/data_processing/augmentation/aug_image.py

class ImageAugmentation():
    """ The Image Augmentation class performs diverse augmentation methods on given
        numpy array. The class acts as an easy to use function/interface for applying
        all types of augmentations with just one function call.

    The class can be configured beforehand by selecting desired augmentation techniques
    and method ranges or strength.
    Afterwards, the class is passed to the [DataGenerator][aucmedi.data_processing.data_generator.DataGenerator]
    which utilizes it during batch generation.

    The specific configurations of selected methods can be adjusted by class variables.

    ???+ abstract "Build on top of the library"
        Albumentations - https://github.com/albumentations-team/albumentations
    """
    #-----------------------------------------------------#
    #              Augmentation Configuration             #
    #-----------------------------------------------------#
    # Define augmentation operator
    operator = None
    # Option for augmentation refinement (padding, cropping and clipping)
    refine = True
    # Augmentation: Flip
    aug_flip = False
    aug_flip_p = 0.5
    # Augmentation: 90 degree rotate
    aug_rotate = False
    aug_rotate_p = 0.5
    # Augmentation: Brightness
    aug_brightness = False
    aug_brightness_p = 0.5
    aug_brightness_limits = (-0.1, 0.1)
    # Augmentation: Contrast
    aug_contrast = False
    aug_contrast_p = 0.5
    aug_contrast_limits = (-0.1, 0.1)
    # Augmentation: Saturation shift
    aug_saturation = False
    aug_saturation_p = 0.5
    aug_saturation_limits = 0.1
    # Augmentation: Hue shift
    aug_hue = False
    aug_hue_p = 0.5
    aug_hue_limits = (-0.1, 0.1)
    # Augmentation: Scale
    aug_scale = False
    aug_scale_p = 0.5
    aug_scale_limits = (0.9, 1.1)
    # Augmentation: Crop
    aug_crop = False
    aug_crop_p = 0.5
    aug_crop_shape = (244, 244)
    # Augmentation: Grid Distortion
    aug_gridDistortion = False
    aug_gridDistortion_p = 0.5
    # Augmentation: Image Compression (JPEG)
    aug_compression = False
    aug_compression_p = 0.5
    aug_compression_limits = (90, 100)
    # Augmentation: Gaussian Noise
    aug_gaussianNoise = False
    aug_gaussianNoise_p = 0.5
    # Augmentation: Gaussian Blur
    aug_gaussianBlur = False
    aug_gaussianBlur_p = 0.5
    # Augmentation: Downscale
    aug_downscaling = False
    aug_downscaling_p = 0.5
    aug_downscaling_effect = 0.10
    # Augmentation: Gamma
    aug_gamma = False
    aug_gamma_p = 0.5
    aug_gamma_limit = (90, 110)
    # Augmentation: Elastic Transformation
    aug_elasticTransform = False
    aug_elasticTransform_p = 0.5

    #-----------------------------------------------------#
    #                    Initialization                   #
    #-----------------------------------------------------#
    def __init__(self, flip=True, rotate=True, brightness=True, contrast=True,
                 saturation=True, hue=True, scale=True, crop=False,
                 grid_distortion=False, compression=False, gaussian_noise=False,
                 gaussian_blur=False, downscaling=False, gamma=False,
                 elastic_transform=False):
        """ Initialization function for the Image Augmentation interface.

        With boolean switches, it is possible to selected desired augmentation techniques.
        Recommended augmentation configurations are defined as class variables.
        Of course, these configs can be adjusted if needed.

        Args:
            flip (bool):                    Boolean, whether flipping should be performed as data augmentation.
            rotate (bool):                  Boolean, whether rotations should be performed as data augmentation.
            brightness (bool):              Boolean, whether brightness changes should be added as data augmentation.
            contrast (bool):                Boolean, whether contrast changes should be added as data augmentation.
            saturation (bool):              Boolean, whether saturation changes should be added as data augmentation.
            hue (bool):                     Boolean, whether hue changes should be added as data augmentation.
            scale (bool):                   Boolean, whether scaling should be performed as data augmentation.
            crop (bool):                    Boolean, whether scaling cropping be performed as data augmentation.
            grid_distortion (bool):         Boolean, whether grid_distortion should be performed as data augmentation.
            compression (bool):             Boolean, whether compression should be performed as data augmentation.
            gaussian_noise (bool):          Boolean, whether gaussian noise should be added as data augmentation.
            gaussian_blur (bool):           Boolean, whether gaussian blur should be added as data augmentation.
            downscaling (bool):             Boolean, whether downscaling should be added as data augmentation.
            gamma (bool):                   Boolean, whether gamma changes should be added as data augmentation.
            elastic_transform (bool):       Boolean, whether elastic deformation should be performed as data augmentation.

        !!! warning
            If class variables (attributes) are modified, the internal augmentation operator
            has to be rebuilt via the following call:

            ```python
            # initialize
            aug = ImageAugmentation(flip=True)

            # set probability to 100% = always
            aug.aug_flip_p = 1.0
            # rebuild
            aug.build()
            ```

        Attributes:
            refine (bool):                  Boolean, whether clipping to [0,255] and padding/cropping should be performed if outside of range.
            aug_flip_p (float):             Probability of flipping application if activated. Default=0.5.
            aug_rotate_p (float):           Probability of rotation application if activated. Default=0.5.
            aug_brightness_p (float):       Probability of brightness application if activated. Default=0.5.
            aug_contrast_p (float):         Probability of contrast application if activated. Default=0.5.
            aug_saturation_p (float):       Probability of saturation application if activated. Default=0.5.
            aug_hue_p (float):              Probability of hue application if activated. Default=0.5.
            aug_scale_p (float):            Probability of scaling application if activated. Default=0.5.
            aug_crop_p (float):             Probability of crop application if activated. Default=0.5.
            aug_grid_distortion_p (float):  Probability of grid_distortion application if activated. Default=0.5.
            aug_compression_p (float):      Probability of compression application if activated. Default=0.5.
            aug_gaussianNoise_p (float):    Probability of gaussian noise application if activated. Default=0.5.
            aug_gaussianBlur_p (float):     Probability of gaussian blur application if activated. Default=0.5.
            aug_downscaling_p (float):      Probability of downscaling application if activated. Default=0.5.
            aug_gamma_p (float):            Probability of gamma application if activated. Default=0.5.
            aug_elasticTransform_p (float): Probability of elastic deformation application if activated. Default=0.5.
        """
        # Cache class variables
        self.aug_flip = flip
        self.aug_rotate = rotate
        self.aug_brightness = brightness
        self.aug_contrast = contrast
        self.aug_scale = scale
        self.aug_crop = crop
        self.aug_saturation = saturation
        self.aug_hue = hue
        self.aug_compression = compression
        self.aug_gaussianNoise = gaussian_noise
        self.aug_gaussianBlur= gaussian_blur
        self.aug_downscaling = downscaling
        self.aug_gamma = gamma
        self.aug_gridDistortion = grid_distortion
        self.aug_elasticTransform = elastic_transform
        # Build augmentation operator
        self.build()

    #-----------------------------------------------------#
    #                Albumentations Builder               #
    #-----------------------------------------------------#
    def build(self):
        """ Builds the albumenations augmentator by initializing  all transformations.

        The activated transformation and their configurations are defined as
        class variables.

        -> Builds a new self.operator
        """
        # Initialize transform list
        transforms = []
        # Fill transform list
        if self.aug_flip:
            tf = ai.Flip(p=self.aug_flip_p)
            transforms.append(tf)
        if self.aug_rotate:
            tf = ai.RandomRotate90(p=self.aug_rotate_p)
            transforms.append(tf)
        if self.aug_brightness:
            tf = ai.RandomBrightnessContrast(brightness_limit=self.aug_brightness_limits,
                                             contrast_limit=0,
                                             p=self.aug_brightness_p)
            transforms.append(tf)
        if self.aug_contrast:
            tf = ai.RandomBrightnessContrast(contrast_limit=self.aug_contrast_limits,
                                             brightness_limit=0,
                                             p=self.aug_contrast_p)
            transforms.append(tf)
        if self.aug_saturation:
            tf = ai.ColorJitter(brightness=0, contrast=0, hue=0,
                                saturation=self.aug_saturation_limits,
                                p=self.aug_saturation_p)
            transforms.append(tf)
        if self.aug_hue:
            tf = ai.ColorJitter(brightness=0, contrast=0, saturation=0,
                                hue=self.aug_hue_limits,
                                p=self.aug_hue_p)
            transforms.append(tf)
        if self.aug_scale:
            tf = ai.RandomScale(scale_limit=self.aug_scale_limits,
                                p=self.aug_scale_p)
            transforms.append(tf)
        if self.aug_crop:
            tf = ai.RandomCrop(width=self.aug_crop_shape[0],
                               height=self.aug_crop_shape[1],
                               p=self.aug_crop_p)
            transforms.append(tf)
        if self.aug_gridDistortion:
            tf = ai.GridDistortion(p=self.aug_gridDistortion_p)
            transforms.append(tf)
        if self.aug_compression:
            tf = ai.ImageCompression(quality_lower=self.aug_compression_limits[0],
                                     quality_upper=self.aug_compression_limits[1],
                                     p=self.aug_compression_p)
            transforms.append(tf)
        if self.aug_gaussianNoise:
            tf = ai.GaussNoise(p=self.aug_gaussianNoise_p)
            transforms.append(tf)
        if self.aug_gaussianBlur:
            tf = ai.GlassBlur(p=self.aug_gaussianBlur_p)
            transforms.append(tf)
        if self.aug_downscaling:
            tf = ai.Downscale(scale_min=self.aug_downscaling_effect,
                              scale_max=self.aug_downscaling_effect,
                              p=self.aug_downscaling_p)
            transforms.append(tf)
        if self.aug_gamma:
            tf = ai.RandomGamma(gamma_limit=self.aug_gamma_limit,
                                p=self.aug_gamma_p)
            transforms.append(tf)
        if self.aug_elasticTransform:
            tf = ai.ElasticTransform(p=self.aug_elasticTransform_p)
            transforms.append(tf)

        # Compose transforms
        self.operator = Compose(transforms)

    #-----------------------------------------------------#
    #                 Perform Augmentation                #
    #-----------------------------------------------------#
    def apply(self, image):
        """ Performs image augmentation with defined configuration on an image.

        This **internal** function is called in the DataGenerator during batch generation.

        Args:
            image (numpy.ndarray):          An image encoded as NumPy array with shape (x, y, channels).
        Returns:
            aug_image (numpy.ndarray):      An augmented / transformed image.
        """
        # Verify that image is in grayscale/RGB encoding
        if np.min(image) < 0 or np.max(image) > 255:
            warnings.warn("Image Augmentation: A value of the image is lower than 0 or higher than 255.",
                          "Albumentations expects images to be in grayscale/RGB!",
                          np.min(image), np.max(image))
        # Cache image shape
        org_shape = image.shape
        # Perform image augmentation
        aug_image = self.operator(image=image)["image"]
        # Perform padding & cropping if image shape changed
        if self.refine and aug_image.shape != org_shape:
            aug_image = ai.pad(aug_image, org_shape[0], org_shape[1], border_mode=cv2.BORDER_REPLICATE, 
                                value=0)
            aug_image = ai.RandomCrop(height=org_shape[0], width=org_shape[1])(image=aug_image)["image"]
        # Perform clipping if image is out of grayscale/RGB encodings
        if self.refine and (np.min(aug_image) < 0 or np.max(aug_image) > 255):
            aug_image = np.clip(aug_image, a_min=0, a_max=255)
        # Return augmented image
        return aug_image

__init__(flip=True, rotate=True, brightness=True, contrast=True, saturation=True, hue=True, scale=True, crop=False, grid_distortion=False, compression=False, gaussian_noise=False, gaussian_blur=False, downscaling=False, gamma=False, elastic_transform=False)

Initialization function for the Image Augmentation interface.

With boolean switches, it is possible to selected desired augmentation techniques. Recommended augmentation configurations are defined as class variables. Of course, these configs can be adjusted if needed.

Parameters:

Name	Type	Description	Default
flip	bool	Boolean, whether flipping should be performed as data augmentation.	True
rotate	bool	Boolean, whether rotations should be performed as data augmentation.	True
brightness	bool	Boolean, whether brightness changes should be added as data augmentation.	True
contrast	bool	Boolean, whether contrast changes should be added as data augmentation.	True
saturation	bool	Boolean, whether saturation changes should be added as data augmentation.	True
hue	bool	Boolean, whether hue changes should be added as data augmentation.	True
scale	bool	Boolean, whether scaling should be performed as data augmentation.	True
crop	bool	Boolean, whether scaling cropping be performed as data augmentation.	False
grid_distortion	bool	Boolean, whether grid_distortion should be performed as data augmentation.	False
compression	bool	Boolean, whether compression should be performed as data augmentation.	False
gaussian_noise	bool	Boolean, whether gaussian noise should be added as data augmentation.	False
gaussian_blur	bool	Boolean, whether gaussian blur should be added as data augmentation.	False
downscaling	bool	Boolean, whether downscaling should be added as data augmentation.	False
gamma	bool	Boolean, whether gamma changes should be added as data augmentation.	False
elastic_transform	bool	Boolean, whether elastic deformation should be performed as data augmentation.	False

Warning

If class variables (attributes) are modified, the internal augmentation operator has to be rebuilt via the following call:

# initialize
aug = ImageAugmentation(flip=True)

# set probability to 100% = always
aug.aug_flip_p = 1.0
# rebuild
aug.build()

Attributes:

Name	Type	Description
refine	bool	Boolean, whether clipping to [0,255] and padding/cropping should be performed if outside of range.
aug_flip_p	float	Probability of flipping application if activated. Default=0.5.
aug_rotate_p	float	Probability of rotation application if activated. Default=0.5.
aug_brightness_p	float	Probability of brightness application if activated. Default=0.5.
aug_contrast_p	float	Probability of contrast application if activated. Default=0.5.
aug_saturation_p	float	Probability of saturation application if activated. Default=0.5.
aug_hue_p	float	Probability of hue application if activated. Default=0.5.
aug_scale_p	float	Probability of scaling application if activated. Default=0.5.
aug_crop_p	float	Probability of crop application if activated. Default=0.5.
aug_grid_distortion_p	float	Probability of grid_distortion application if activated. Default=0.5.
aug_compression_p	float	Probability of compression application if activated. Default=0.5.
aug_gaussianNoise_p	float	Probability of gaussian noise application if activated. Default=0.5.
aug_gaussianBlur_p	float	Probability of gaussian blur application if activated. Default=0.5.
aug_downscaling_p	float	Probability of downscaling application if activated. Default=0.5.
aug_gamma_p	float	Probability of gamma application if activated. Default=0.5.
aug_elasticTransform_p	float	Probability of elastic deformation application if activated. Default=0.5.

Source code in aucmedi/data_processing/augmentation/aug_image.py

def __init__(self, flip=True, rotate=True, brightness=True, contrast=True,
             saturation=True, hue=True, scale=True, crop=False,
             grid_distortion=False, compression=False, gaussian_noise=False,
             gaussian_blur=False, downscaling=False, gamma=False,
             elastic_transform=False):
    """ Initialization function for the Image Augmentation interface.

    With boolean switches, it is possible to selected desired augmentation techniques.
    Recommended augmentation configurations are defined as class variables.
    Of course, these configs can be adjusted if needed.

    Args:
        flip (bool):                    Boolean, whether flipping should be performed as data augmentation.
        rotate (bool):                  Boolean, whether rotations should be performed as data augmentation.
        brightness (bool):              Boolean, whether brightness changes should be added as data augmentation.
        contrast (bool):                Boolean, whether contrast changes should be added as data augmentation.
        saturation (bool):              Boolean, whether saturation changes should be added as data augmentation.
        hue (bool):                     Boolean, whether hue changes should be added as data augmentation.
        scale (bool):                   Boolean, whether scaling should be performed as data augmentation.
        crop (bool):                    Boolean, whether scaling cropping be performed as data augmentation.
        grid_distortion (bool):         Boolean, whether grid_distortion should be performed as data augmentation.
        compression (bool):             Boolean, whether compression should be performed as data augmentation.
        gaussian_noise (bool):          Boolean, whether gaussian noise should be added as data augmentation.
        gaussian_blur (bool):           Boolean, whether gaussian blur should be added as data augmentation.
        downscaling (bool):             Boolean, whether downscaling should be added as data augmentation.
        gamma (bool):                   Boolean, whether gamma changes should be added as data augmentation.
        elastic_transform (bool):       Boolean, whether elastic deformation should be performed as data augmentation.

    !!! warning
        If class variables (attributes) are modified, the internal augmentation operator
        has to be rebuilt via the following call:

        ```python
        # initialize
        aug = ImageAugmentation(flip=True)

        # set probability to 100% = always
        aug.aug_flip_p = 1.0
        # rebuild
        aug.build()
        ```

    Attributes:
        refine (bool):                  Boolean, whether clipping to [0,255] and padding/cropping should be performed if outside of range.
        aug_flip_p (float):             Probability of flipping application if activated. Default=0.5.
        aug_rotate_p (float):           Probability of rotation application if activated. Default=0.5.
        aug_brightness_p (float):       Probability of brightness application if activated. Default=0.5.
        aug_contrast_p (float):         Probability of contrast application if activated. Default=0.5.
        aug_saturation_p (float):       Probability of saturation application if activated. Default=0.5.
        aug_hue_p (float):              Probability of hue application if activated. Default=0.5.
        aug_scale_p (float):            Probability of scaling application if activated. Default=0.5.
        aug_crop_p (float):             Probability of crop application if activated. Default=0.5.
        aug_grid_distortion_p (float):  Probability of grid_distortion application if activated. Default=0.5.
        aug_compression_p (float):      Probability of compression application if activated. Default=0.5.
        aug_gaussianNoise_p (float):    Probability of gaussian noise application if activated. Default=0.5.
        aug_gaussianBlur_p (float):     Probability of gaussian blur application if activated. Default=0.5.
        aug_downscaling_p (float):      Probability of downscaling application if activated. Default=0.5.
        aug_gamma_p (float):            Probability of gamma application if activated. Default=0.5.
        aug_elasticTransform_p (float): Probability of elastic deformation application if activated. Default=0.5.
    """
    # Cache class variables
    self.aug_flip = flip
    self.aug_rotate = rotate
    self.aug_brightness = brightness
    self.aug_contrast = contrast
    self.aug_scale = scale
    self.aug_crop = crop
    self.aug_saturation = saturation
    self.aug_hue = hue
    self.aug_compression = compression
    self.aug_gaussianNoise = gaussian_noise
    self.aug_gaussianBlur= gaussian_blur
    self.aug_downscaling = downscaling
    self.aug_gamma = gamma
    self.aug_gridDistortion = grid_distortion
    self.aug_elasticTransform = elastic_transform
    # Build augmentation operator
    self.build()

apply(image)

Performs image augmentation with defined configuration on an image.

This internal function is called in the DataGenerator during batch generation.

Parameters:

Name	Type	Description	Default
image	numpy.ndarray	An image encoded as NumPy array with shape (x, y, channels).	required

Returns:

Name	Type	Description
aug_image	numpy.ndarray	An augmented / transformed image.

Source code in aucmedi/data_processing/augmentation/aug_image.py

def apply(self, image):
    """ Performs image augmentation with defined configuration on an image.

    This **internal** function is called in the DataGenerator during batch generation.

    Args:
        image (numpy.ndarray):          An image encoded as NumPy array with shape (x, y, channels).
    Returns:
        aug_image (numpy.ndarray):      An augmented / transformed image.
    """
    # Verify that image is in grayscale/RGB encoding
    if np.min(image) < 0 or np.max(image) > 255:
        warnings.warn("Image Augmentation: A value of the image is lower than 0 or higher than 255.",
                      "Albumentations expects images to be in grayscale/RGB!",
                      np.min(image), np.max(image))
    # Cache image shape
    org_shape = image.shape
    # Perform image augmentation
    aug_image = self.operator(image=image)["image"]
    # Perform padding & cropping if image shape changed
    if self.refine and aug_image.shape != org_shape:
        aug_image = ai.pad(aug_image, org_shape[0], org_shape[1], border_mode=cv2.BORDER_REPLICATE, 
                            value=0)
        aug_image = ai.RandomCrop(height=org_shape[0], width=org_shape[1])(image=aug_image)["image"]
    # Perform clipping if image is out of grayscale/RGB encodings
    if self.refine and (np.min(aug_image) < 0 or np.max(aug_image) > 255):
        aug_image = np.clip(aug_image, a_min=0, a_max=255)
    # Return augmented image
    return aug_image

build()

Builds the albumenations augmentator by initializing all transformations.

The activated transformation and their configurations are defined as class variables.

-> Builds a new self.operator

Source code in aucmedi/data_processing/augmentation/aug_image.py

def build(self):
    """ Builds the albumenations augmentator by initializing  all transformations.

    The activated transformation and their configurations are defined as
    class variables.

    -> Builds a new self.operator
    """
    # Initialize transform list
    transforms = []
    # Fill transform list
    if self.aug_flip:
        tf = ai.Flip(p=self.aug_flip_p)
        transforms.append(tf)
    if self.aug_rotate:
        tf = ai.RandomRotate90(p=self.aug_rotate_p)
        transforms.append(tf)
    if self.aug_brightness:
        tf = ai.RandomBrightnessContrast(brightness_limit=self.aug_brightness_limits,
                                         contrast_limit=0,
                                         p=self.aug_brightness_p)
        transforms.append(tf)
    if self.aug_contrast:
        tf = ai.RandomBrightnessContrast(contrast_limit=self.aug_contrast_limits,
                                         brightness_limit=0,
                                         p=self.aug_contrast_p)
        transforms.append(tf)
    if self.aug_saturation:
        tf = ai.ColorJitter(brightness=0, contrast=0, hue=0,
                            saturation=self.aug_saturation_limits,
                            p=self.aug_saturation_p)
        transforms.append(tf)
    if self.aug_hue:
        tf = ai.ColorJitter(brightness=0, contrast=0, saturation=0,
                            hue=self.aug_hue_limits,
                            p=self.aug_hue_p)
        transforms.append(tf)
    if self.aug_scale:
        tf = ai.RandomScale(scale_limit=self.aug_scale_limits,
                            p=self.aug_scale_p)
        transforms.append(tf)
    if self.aug_crop:
        tf = ai.RandomCrop(width=self.aug_crop_shape[0],
                           height=self.aug_crop_shape[1],
                           p=self.aug_crop_p)
        transforms.append(tf)
    if self.aug_gridDistortion:
        tf = ai.GridDistortion(p=self.aug_gridDistortion_p)
        transforms.append(tf)
    if self.aug_compression:
        tf = ai.ImageCompression(quality_lower=self.aug_compression_limits[0],
                                 quality_upper=self.aug_compression_limits[1],
                                 p=self.aug_compression_p)
        transforms.append(tf)
    if self.aug_gaussianNoise:
        tf = ai.GaussNoise(p=self.aug_gaussianNoise_p)
        transforms.append(tf)
    if self.aug_gaussianBlur:
        tf = ai.GlassBlur(p=self.aug_gaussianBlur_p)
        transforms.append(tf)
    if self.aug_downscaling:
        tf = ai.Downscale(scale_min=self.aug_downscaling_effect,
                          scale_max=self.aug_downscaling_effect,
                          p=self.aug_downscaling_p)
        transforms.append(tf)
    if self.aug_gamma:
        tf = ai.RandomGamma(gamma_limit=self.aug_gamma_limit,
                            p=self.aug_gamma_p)
        transforms.append(tf)
    if self.aug_elasticTransform:
        tf = ai.ElasticTransform(p=self.aug_elasticTransform_p)
        transforms.append(tf)

    # Compose transforms
    self.operator = Compose(transforms)