DataProcessing

class DataProcessing(samples, dim=2, **kwargs)

Bases: object

Process multiple SingleImages

See also

SingleImage2D

static _get_files(directory, extensions)

return files with extensions

Parameters:

• directory (str) – directory containing the files
• extensions (list) – list of strings specifying valid extensions

Returns:

valid files

Return type:

list

classmethod from_dir(data_dir, verbose=True, n_jobs=None, n_dim=2)

create class instance from directory

Parameters:

• data_dir (str) – directory where shapedata is stored
• verbose (bool) – whether or not to print current progress
• n_jobs (int) – number of jobs for loading data (default: None -> all available CPUs are used)
• n_dim (int) – Integer indicating the dimensionality of the image (default: 2)

Returns:

class instance

Return type:

DataProcessing

images

get list of samples’ pixels

Returns:pixels Return type:list

landmarks

get list of samples’ landmarks

Returns:landmarks Return type:list

lmk_pca(scale: bool, center: bool, pt_indices=[], *args, **kwargs)

perform PCA on samples’ landmarks

Parameters:

• scale (bool) – whether or not to scale the principal components with the corresponding eigen value
• center (bool) – whether or not to substract mean before pca
• pt_indices (int) – indices to include into PCA (if empty: include all points)
• args (list) – additional positional arguments (passed to pca)
• **kwargs – additional keyword arguments (passed to pca)

Returns:

eigen_shapes

Return type:

np.array

resize(img_size)

resize all samples

Parameters:img_size (tuple) – new image size

SingleImage2D

class SingleImage2D(img, lmk=None, **kwargs)

Bases: shapedata.base_data_processing.BaseSingleImage

Holds Single Image

_crop(min_y, min_x, max_y, max_x)

Implements actual cropping inplace

Parameters:

• min_y (int) – minimum y value
• min_x (int) – minimum x value
• max_y (int) – maximum y value
• max_x (int) – maximum x value

Returns:

cropped image

Return type:

SingleImage2D

static _crop_lmks(lmks, min_y, min_x, max_y, max_x)

Crops landmarks to given values

Parameters:

• lmks (np.ndarray) – landmarks to crop
• min_y (int) – minimum y value
• min_x (int) – minimum x value
• max_y (int) – maximum y value
• max_x (int) – maximum x value

Returns:

cropped landmarks

Return type:

np.ndarray

_crop_to_landmarks(proportion=0.0, **kwargs)

Crop to landmarks

Parameters:

• proportion (float) – Cropping Proportion
• **kwargs – additional keyword arguments (ignored here)

Returns:

Cropped Image

Return type:

SingleImage2D

_normalize_rotation(lmks, index_left, index_right, **kwargs)

normalizes rotation based on two keypoints

Parameters:

• lmks (np.ndarray) – landmarks for rotation normalization
• index_left (int) – index for left point
• index_right (int) – index for right point
• **kwargs – additional keyword arguments (passed to skimage.transform.warp())

Returns:

transformed Image

Return type:

BaseSingleImage

_save_landmarks(filepath, lmk_type, **kwargs)

Saves landmarks to file

Parameters:

• filepath (str) – path to file the landmarks should be saved to
• lmk_type (str) – specifies the type of landmark file
• **kwargs – additional keyword arguments passed to save function

Raises:

ValueError – no valid landmarktype is given

_transform_img(transformation: <sphinx.ext.autodoc.importer._MockObject object at 0x7f8f65aab438>, **kwargs)

Apply transformation inplace to image

Parameters:

• transformation (skimage.transform.AffineTransform) – transformation to apply
• **kwargs – additional keyword arguments

Returns:

Transformed Image with original Landmarks

Return type:

BaseSingleImage

_transform_lmk(transformation: <sphinx.ext.autodoc.importer._MockObject object at 0x7f8f65af4748>)

Apply transformation inplace to landmarks

Parameters:transformation (skimage.transform.AffineTransform) – transformation to apply Returns:Image with Transformed Landmarks Return type:BaseSingleImage

apply_trafo(transformation: <sphinx.ext.autodoc.importer._MockObject object at 0x7f8f65aab438>, **kwargs)

Apply transformation inplace to image and landmarks

Parameters:

• transformation (skimage.transform.AffineTransform) – transformation to apply
• **kwargs – additional keyword arguments

Returns:

Transformed Image and Landmarks

Return type:

BaseSingleImage

cartesian_coordinates()

Transforms landmark coordinates inplace to cartesian coordinates

Returns:class – Image with Landmarks in cartesian Coordinates Return type:SingleImage2D

crop(min_y, min_x, max_y, max_x)

Crops Image by specified values

Parameters:

• min_y (int) – minimum y value
• min_x (int) – minimum x value
• max_y (int) – maximum y value
• max_x (int) – maximum x value

Returns:

cropped image

Return type:

BaseSingleImage

crop_to_landmarks(proportion=0.0, **kwargs)

Crop image to landmarks

Parameters:

• proportion (float) – image proportion to add to size of bounding box
• **kwargs – additional keyword arguments

Returns:

cropped image

Return type:

BaseSingleImage

classmethod from_files(file, extension=None, **kwargs)

Creates a class instance from files

Parameters:

• *args – positional arguments
• **kwargs – keyword arguments

Raises:

NotImplementedError

classmethod from_ljson_files(img_file, **kwargs)

Creates class from menpo pts landmarks and image

Parameters:

• img_file (str) – image file to load
• **kwargs – additional keyword arguments

Returns:

class instance

Return type:

SingleImage2D

classmethod from_npy_files(file, **kwargs)

Create class from image file :param file: path to image file :type file: str :param **kwargs: additional keyword arguments

Returns:class instance Return type:SingleImage2D

classmethod from_pts_files(img_file, **kwargs)

Creates class from menpo ljson landmarks and image

Parameters:

• img_file (str) – image file to load
• **kwargs – additional keyword arguments

Returns:

class instance

Return type:

SingleImage2D

static get_landmark_bounds(lmks)

Function to calculate the landmark bounds

Parameters:lmks (np.ndarray) – landmarks Returns:

• int (min_y)
• int (min_x)
• int (max_y)
• int (max_x)

homogeneous_coordinates()

Transforms landmark coordinates inplace to homogeneous coordinates

Returns:Image with Landmarks in Homogeneous Coordinates Return type:SingleImage2D

img

Property to get the actual image pixels

Returns:image pixels Return type:np.array

is_gray

Property returning whether the image is a grayscale image

Raises:NotImplementedError – if not overwritten by subclass

is_homogeneous

Property returning whether the landmarks are in homogeneous coordinates

Raises:NotImplementedError – if not overwritten by subclass

normalize_rotation(index_left, index_right, **kwargs)

normalizes rotation based on two keypoints

index_left : int

landmark-index of the left point

index_right : int

landmark-index of the right point

**kwargs:

additional keyword arguments (passed to warp())

Returns:normalized image Return type:SingleImage2D

rescale(scale, **kwargs)

Scale Image and landmarks

Parameters:

• scale – scale parameter
• **kwargs – additional keyword arguments (passed to skimage.transform.warp())

Returns:

transformed Image

Return type:

BaseSingleImage

resize(target_shape, **kwargs)

resize image and scale landmarks :param target_shape: target shape for resizing :type target_shape: tuple or list :param **kwargs: additional keyword arguments (passed to

skimage.transform.warp())

Returns:transformed Image Return type:BaseSingleImage

rotate(angle, degree=True, **kwargs)

Rotates the image and landmarks by given angle

Parameters:

• angle (float or int) – rotation angle
• degree (bool) – whether the angle is given in degree or radiant
• **kwargs – additional keyword arguments (passed to skimage.transform.warp())

Returns:

transformed Image

Return type:

BaseSingleImage

save(directory, filename, lmk_type='LJSON', **kwargs)

Saves Image and optionally landmarks to files

Parameters:

• directory (str) – string containing the directory to save
• filename (str) – string containing the filename (without the extension)
• lmk_type (str or None) – if None: no landmarks will be saved if str: specifies type of landmark file
• **kwargs – additional keyword arguments passed to save function for landmarks

save_image(filepath)

Saves Image to file

Parameters:filepath (str) – file to save the image to

save_landmarks(filepath, lmk_type='LJSON', **kwargs)

Saves landmarks to file

Parameters:

• filepath (str) – path to file the landmarks should be saved to
• lmk_type (str) – specifies the type of landmark file
• **kwargs – additional keyword arguments passed to save function

to_grayscale()

Convert Image to grayscale

Returns:Grayscale Image Return type:BaseSingleImage

transform(transform=None, rotation=None, scale=None, translation=None, shear=None, trafo_matrix=None, return_matrix=False, **kwargs)

transform image and landmarks by parameters or transformation matrix See skimage.transform.AffineTransform for a detailed parameter explanation

Parameters:

• transform (skimage.transform.AffineTransform) – if transform is specified it overwrites all other arguments
• rotation (float or None) – rotation angle in radiant
• scale (float or None) – scale value
• translation – translation params
• shear – shear params
• trafo_matrix – transformation matrix
• return_matrix (bool) – whether to return the transformation matrix along the transformed object
• **kwargs – additional keyword arguments

Returns:

• BaseSingleImage – transformed Image
• [optional] np.ndarray – transformation matrix

transform_about_centre(transform=None, rotation=None, scale=None, translation=None, shear=None, trafo_matrix=None, return_matrix=False, **kwargs)

Perform transformations about the image center. (internally shifting image to origin, perform transformation and shift it back)

Parameters:

• transform (skimage.transform.AffineTransform) – if transform is specified it overwrites all other arguments
• rotation (float) – rotation angle in radiant
• scale (float) – scale value
• translation – translation params
• shear – shear params
• trafo_matrix – transformation matrix
• return_matrix (bool) – whether to return the transformation matrix along the transformed object
• **kwargs – additional keyword arguments

Returns:

• BaseSingleImage – transformed Image
• [optional] np.ndarray – transformation matrix

translate(translation, relative=False, **kwargs)

translates image and landmarks

translation :

translation parameters

relative : bool

whether translation parameters are relative to image size

**kwargs :

additional keyword arguments (passed to skimage.transform.warp())

BaseSingleImage

transformed Image

view(view_landmarks=False, create_fig=False, **kwargs)

Shows image (and optional the landmarks)

Parameters:

• view_landmarks (bool) – whether or not to show the landmarks
• **kwargs – additional keyword arguments (are passed to imshow)

Returns:

figure with plot

Return type:

Figure