FASSEGHandle

class hybrid_learning.datasets.custom.fasseg.FASSEGHandle(dataset_root, annotations_root=None, part=None, part_name=None, **kwargs)

Bases: BaseDataset

Handle for FASSEG-like datasets.

Note

The original FASSEG dataset is not required for this handle. Any dataset with a format following that of the FASSEG dataset is supported (folder structure, file types, color codes).

The FASSEG dataset can be found here:: https://github.com/massimomauro/FASSEG-repository

The required structure for supported datasets is very simple

• under dataset_root all input images

• under annotations_root all annotation files with the same file name as the corresponding input image

Once a handle is instantiated, its __getitem__ yields tuples of input image and binary part segmentation mask (possibly transformed by hybrid_learning.datasets.base.BaseDataset.transforms). For details see getitem().

Public Data Attributes:

part_name

The string name of the part that is extracted from the masks.

Inherited from : py: class:BaseDataset

settings

Settings of the instance.

Public Methods:

getitem(i)	Load image and its mask at index i and select binary part mask.
descriptor(i)	Return the image file name for index i.
image_filepath(i)	Provide the path to the image at index i.
mask_filepath(i)	Provide the path to the mask at index i.

Inherited from : py: class:BaseDataset

getitem(i)	Load image and its mask at index i and select binary part mask.
descriptor(i)	Return the image file name for index i.

Special Methods:

__init__(dataset_root[, annotations_root, ...])	Init.
__len__()	Number of data points in the dataset; to be implemented in subclasses.

Inherited from : py: class:BaseDataset

__init__(dataset_root[, annotations_root, ...])	Init.
__len__()	Number of data points in the dataset; to be implemented in subclasses.
__getitem__(idx)	Get item from idx in dataset with transformations applied.
__repr__()	Nice printing function.

Inherited from : py: class:Dataset

__getitem__(idx)	Get item from idx in dataset with transformations applied.
__add__(other)

---

__init__(dataset_root, annotations_root=None, part=None, part_name=None, **kwargs)

Init.

Parameters

• dataset_root (str) – the directory under which to find the images;

• annotations_root (Optional[str]) –

  the directory under which to find the segmentation masks; assumes as default that

  • dataset_root is path/split_RGB and

  • annotations_root is path/split_Labels

• part (Optional[Union[FASSEGParts, Tuple[int, int, int]]]) – the FASSEGParts instance to use the color of or the color tuple

• part_name (Optional[str]) – the string name of the part that is extracted; overridden by part.name if part features such an attribute

• kwargs – parameters for BaseDataset

__len__()

Number of data points in the dataset; to be implemented in subclasses.

descriptor(i)

Return the image file name for index i. This is unique throughout a FASSEG like dataset and can be used for e.g. image IDs for caching.

Parameters

i (int) –

Return type

str

getitem(i)

Load image and its mask at index i and select binary part mask.

Used for __getitem__(). The value of the return tuple (input_img, part_mask) are torch.Tensor representations of PIL.Image.Image image instances.

Parameters

i (int) –

Return type

Tuple[Image, Image]

image_filepath(i)

Provide the path to the image at index i.

mask_filepath(i)

Provide the path to the mask at index i.

__parameters__ = ()

annotations_root: str

Path to the annotations root folder under which to find the annotation files.

img_fns: List[str]

List of file names of images (and their annotations) handled by this instance. Images can be found in the dataset_root and annotations in annotations_root. These are used for getitem().

part

Part of the face and its color to select mask of.

property part_name: Optional[str]

The string name of the part that is extracted from the masks.