Python API¶

Database¶

class bob.db.iostar.query.FileList(db_json)[source]¶

Bases: object

FileList object that loads the protocol as defined in a json file. Provides a __getitem__ interface.

class bob.db.iostar.query.Database(protocol='default_vessel')[source]¶

Bases: object

A low level database interface to be used with PyTorch or other deep learning frameworks.

protocol¶

protocol defining the train-test split.

Type: str

property paths¶: returns: paths – list of all paths of all samples :rtype: list

samples(split=None)[source]¶

Given a split, returns a list of Sample objects.

Parameters: split (str) – ‘train’, ‘test’ or None (returns all samples)
Returns: samples – list of Sample objects
Return type: list

Sample¶

class bob.db.iostar.models.FundusImage(path)[source]¶

Bases: object

Generic fundus image object.

property basename¶

Returns the file name

Returns: name
Return type: str

property size¶: returns: size – the fundus image resolution in (W, H). :rtype: tuple

pil_image(datadir=None)[source]¶

Returns: img
Return type: PIL.Image.Image

class bob.db.iostar.models.GroundTruth(path, threshold=None, drawsize=None, mask=None)[source]¶

Bases: object

Generic ground truth object.

Allows for thresholding in case there are multiple ground truth annotations in one bitmap
Allows for “on-the-fly” drawing of ground truth annotations with a specified size
Allows for mask removal from gt (special case for IOSTAR)

Parameters

path (str) – relative path to the file on disk
threshold (float) – in range [0,1] used to threshold ground-truth image data with multiple classes e.g. optic disc and cup
drawsize (tuple) – size tuple in pillow (W, H) format. Used for cases where drawing of gt is required
mask (Mask) –

property basename¶

Returns the file name

Returns: name
Return type: str

property size¶

Retirms the ground truth image resolution in (W, H).

Returns: size
Return type: tuple

pil_image(datadir=None)[source]¶

Returns: gt – mode = ‘1’
Return type: PIL.Image.Image

class bob.db.iostar.models.Mask(path, remove=None)[source]¶

Bases: object

Generic mask object.

property basename¶

Returns the file name

Returns: name
Return type: str

property size¶: returns: size – the mask image resolution in (W, H). :rtype: tuple

pil_image(datadir=None)[source]¶

Returns: mask – mode = ‘1’
Return type: PIL.Image.Image

class bob.db.iostar.models.Sample(img, gt, mask)[source]¶

Bases: object

Generic Sample object

High level sample class that combines the objects ‘FundusImage’, ‘GroundTruth’ and ‘Mask’. Allows for access of the subclass, e.g. :

>>> img = FundusImage('path/to/some_img.file')
>>> gt = GroundTruth('path/to/some_gt.file')
>>> mask = Mask('path/to/some_mask.file')
>>> mysample = Sample(img, gt, mask)
>>> mysample.img.basename
'some_img.file'
>>> mysample.gt.basename
'some_gt.file'
>>> mysample.mask.basename
'some_mask.file'
>>>

Parameters

img (FundusImage) –
gt (GroundTruth) –
mask (Mask) –

property paths¶: returns: paths – paths of image, ground truth and mask :rtype: list