cvdatasets/nabirds/dataset

Dataset

The provided Dataset class inherits from the AnnotationsReadMixin and the PartMixing. Consequently, to create a Dataset instance an Annotation object is required (it is necessary to pass the names of the arguments!):

# replace NAB_Annotations with CUB_Annotations to load CUB200-2011 annotations
from nabirds import NAB_Annotations, Dataset

annot = NAB_Annotations("path/to/nab/folder")

train_data = Dataset(uuids=annot.train_uuids, annotations=annot)
test_data = Dataset(uuids=annot.test_uuids, annotations=annot)

Working with part and bounding box annotations

In order to enable the PartMixing functionalities pass the appropriate arguments to the initializer:

data = Dataset(
	uuids=uuids, annotations=annot,

	uniform_parts=True,

	crop_to_bb=True,
	crop_uniform=True,

	parts_in_bb=True,

	rnd_select=True,
	ratio=0.35,
	seed=21486214
)

Information about all possible PartMixing functionalities can be found here.

All of the used part and bounding box functionalities can be manually used with the help of the utils module.

utils module

TBD (see main page for more information for now)