bioimageloader.base#

Define a base class and its interface

Dataset is the base of all datasets

MaskDataset is the base of datasets that have mask annotation

class bioimageloader.base.DatasetInterface[source]#

Dataset interface

Attributes

__repr__: Print info of dataset
__len__
__getitem__: Given index returns a dictionary of key(s) and array
acronym
root_dir: Path to root directory
file_list: A list of pathes to image files

Methods

get_image(key)

Get an image

abstract get_image(key)[source]#: Get an image

abstract class property acronym#: Assign acroym for a subclass dataset

abstract property root_dir#: Path to root directory

property file_list#: A list of pathes to image files

class bioimageloader.base.Dataset[source]#

Base to define common attributes and methods for [MaskDataset, …]

Notes

Required attributes in subclass

anno_dict
__getitem__()
get_image()

Attributes

__repr__: Print info of dataset
__len__: Length of dataset.
__iter__
root_dir: Path to root directory
output: Determine return(s) when called, fixed to ‘image’
transforms: Transform images and masks
num_samples: Number of calls that will override __len__
grayscaleoptional: Flag for grayscale conversion
grayscale_modeoptional: Determine grayscale mode one of {‘cv2’, ‘equal’, Sequence[float]}
num_channelsoptional: Number of image channels used for to_gray()

Methods

`__getitem__`(ind)	Get image
`_drop_missing_pairs`()	Drop images and reindex the anno list (dict)
`to_gray`(arr[, grayscale_mode, num_channels])	Convert bioimage to grayscale

property root_dir: pathlib.Path#: Path to root directory

property output: str#: Determine return(s) when called, fixed to ‘image’

property transforms: Optional[albumentations.core.composition.Compose]#: Transform images and masks

property num_samples: Optional[int]#: Number of calls that will override __len__

property grayscale: Optional[bool]#: Flag for grayscale conversion

property grayscale_mode: Optional[Union[str, Sequence[float]]]#: Determine grayscale mode one of {‘cv2’, ‘equal’, Sequence[float]}

property num_channels: Optional[int]#: Number of image channels used for to_gray()

static to_gray(arr: numpy.ndarray, grayscale_mode: Optional[Union[str, Sequence[float]]] = None, num_channels: int = 3) → numpy.ndarray[source]#

Convert bioimage to grayscale

Parameters

arrimage array: Numpy image array whose shape is (h, w, 3)
grayscale_modestr or sequence of float, optional: Choose a strategy for gray conversion. Three options are availble. Either one of {‘cv2’, ‘equal’} or be a sequence of float numbers, which indicate linear weights of each channel.
num_channelsint: Explicitly set number of channels for grayscale_mode=’equal’.

class bioimageloader.base.MaskDataset[source]#

Base for datasets with mask annotation

Define __getitem__ method to load mask annotation paired with image. Pre-defined attributes are prefixed with a single underscore to distinguish them from those specific to a dataset. It is required to implement two methods: get_image() and get_mask() as well as acronym and _root_dir for each subclass.