Python API

bob.db.celeba.get_config()

Returns a string containing the configuration information.

class bob.db.celeba.Annotation(file_id, labels)

Annotations of the CelebA database consists of the locations of the two eyes, the nose tip and the mouth corners. There is exactly one annotation for each File, which after creation of the database can be obtained using self.file.

Note

The eye and mouth locations given in the original annotation file are in observer perspective (i.e., the right eye is to the right of the left eye). To be consistent with other databases of Bob, here we change them to be in subject perspective (i.e., the right eye is to the left of the left eye). Also, as usual for Bob, coordinates are specified in (y,x) order, opposite to how they are written in the list file.

class bob.db.celeba.Attributes(file_id, attributes)

Attributes of the CelebA database consist of a list of binary labels for the face, such as Smiling, Wavy_Hair, Young or Pointy_Nose. Each attribute is binary, i.e., it either exists (+1) or it does not exist (-1). There is exactly one set of attributes for each File, which after creation of the database can be obtained using self.file.

attribute_indices = {'Sideburns': 30, 'Black_Hair': 8, 'Wavy_Hair': 33, 'Young': 39, 'Heavy_Makeup': 18, 'Blond_Hair': 9, 'Attractive': 2, '5_o_Clock_Shadow': 0, 'Wearing_Necktie': 38, 'Blurry': 10, 'Double_Chin': 14, 'Brown_Hair': 11, 'Mouth_Slightly_Open': 21, 'Goatee': 16, 'Bald': 4, 'Pointy_Nose': 27, 'Gray_Hair': 17, 'Pale_Skin': 26, 'Arched_Eyebrows': 1, 'Wearing_Hat': 35, 'Receding_Hairline': 28, 'Straight_Hair': 32, 'Big_Nose': 7, 'Rosy_Cheeks': 29, 'Oval_Face': 25, 'Bangs': 5, 'Male': 20, 'Mustache': 22, 'High_Cheekbones': 19, 'No_Beard': 24, 'Eyeglasses': 15, 'Bags_Under_Eyes': 3, 'Wearing_Necklace': 37, 'Wearing_Lipstick': 36, 'Big_Lips': 6, 'Narrow_Eyes': 23, 'Chubby': 13, 'Smiling': 31, 'Bushy_Eyebrows': 12, 'Wearing_Earrings': 34}

attribute_names = ('5_o_Clock_Shadow', 'Arched_Eyebrows', 'Attractive', 'Bags_Under_Eyes', 'Bald', 'Bangs', 'Big_Lips', 'Big_Nose', 'Black_Hair', 'Blond_Hair', 'Blurry', 'Brown_Hair', 'Bushy_Eyebrows', 'Chubby', 'Double_Chin', 'Eyeglasses', 'Goatee', 'Gray_Hair', 'Heavy_Makeup', 'High_Cheekbones', 'Male', 'Mouth_Slightly_Open', 'Mustache', 'Narrow_Eyes', 'No_Beard', 'Oval_Face', 'Pale_Skin', 'Pointy_Nose', 'Receding_Hairline', 'Rosy_Cheeks', 'Sideburns', 'Smiling', 'Straight_Hair', 'Wavy_Hair', 'Wearing_Earrings', 'Wearing_Hat', 'Wearing_Lipstick', 'Wearing_Necklace', 'Wearing_Necktie', 'Young')

class bob.db.celeba.Database(original_directory=None, original_extension='.jpg')

Wrapper class for the MNIST database of handwritten digits (http://yann.lecun.com/exdb/mnist/).

Creates the database

annotations(self, file) → annotations

Returns the annotations for the given file.

Note

The annotations are returned in the default way comparable with other bob databases, i.e., in (y,x) order. Please also note that the name of the annotations have been changed to be in subject perspective (e.g., the x-coordinate of the left eye is larger than the x coordintae of the right eye), while the original database contains eye locations in observer perspective.

Parameters:

file : File

The file object to get the annotations for.

Returns:

annotations : {}

The dictionary of annotations, which include the coordinated for ‘reye’, ‘leye’, ‘nose’, ‘rmouth’, ‘lmouth’.

attribute_names()

Returns the vector of labels

attributes(self, file, attribute_names=None) → attributes

Returns the list of attributes for the given file. If attribute_names are specified, only the attributes for the given attribute_names are returned – in exactly the same order.

Parameters:

file : File

The file object to get the attributes for.

attribute_names : str or [str] or None

The list of attribute names (see Attribute.attribute_names for a list of valid names) which should be retrieved. If None, all attributes are returned.

Returns:

attributes : [int]

The list of attributes, which are either +1 (in case the attribute is present) or -1 (in case the attribute is absent).

objects(self, purposes=None, with_attributes=None, without_attributes=None) → files

Returns a list of File objects for the given purposes.

Parameters:

purposes : str or [str] or None

A purpose or a list of purposes, which might be ("training", "validation", "test"), or ("world", "dev", "eval"). If None, the full list of purposes will be used.

with_attributes : str or [str] or None

If given, the resulting list is filtered to contain only objects with the given list of attributes. If several attributes are specified, the are combined with a logical and.

without_attributes : str or [str] or None

If given, the resulting list is filtered to contain only objects that do not possess the given list of attributes. If several attributes are specified, the are combined with a logical and.

Returns:

files : [File]

A list of files for the given purpose(s), where the given attributes are filtered.

original_file_name(self, file) → image_name

Returns the full path of the original image name. This function requires that the original_directory was specified in the constructor of this class.

Parameters:

file : : File

The file object to get the original file name for.

Returns:

image_name : str

The full path to the image belonging to the given File object.

test_set(self) → files

Returns a list of File objects that can be used for testing

Returns:

files : [File]

A list of test files.

training_set(self) → files

Returns a list of File objects that can be used for training

Returns:

files : [File]

A list of training files.

validation_set(self) → files

Returns a list of File objects that can be used for validation purposes

Returns:

files : [File]

A list of validation files.

class bob.db.celeba.File(file_id, purpose)

Information about the files of the CelebA database.

Each file includes only a numerical ID (which is identical to the file path).

Additionally, some fields will be available, once the database is created:

• annotation: the Annotation, which belongs to this file
• attributes: the Attributes, which belongs to this file
• purpose: the Purpose, for which this file is used

make_path(directory=None, extension=None)

Wraps the current path so that a complete path is formed. To get the original file name, please use the Database.original_file_name() function instead, or set the add_sighting_id flag to False.

Keyword parameters:

directory : str or None

An optional directory name that will be prefixed to the returned result.

extension : str or None

An optional extension that will be suffixed to the returned filename. The extension normally includes the leading . character as in .jpg or .hdf5.

Returns a string containing the newly generated file path, which by default is unique.