fuel.datasets.DogsVsCats

Complexity

Total Complexity

1

Size/Duplication

Total Lines	21
Duplicated Lines	0 %

1 Method

Rating	Name	Duplication	Size	Complexity
A	__init__()	0	4	1

from fuel.datasets import H5PYDataset
from fuel.transformers import ScaleAndShift
from fuel.utils import find_in_data_path


class DogsVsCats(H5PYDataset):
    """The Kaggle Dogs vs. Cats dataset of cats and dogs images.

    Parameters
    ----------
    which_sets : tuple of str
        Which split to load. Valid values are 'train', 'valid' and 'test'.
        The 'train' set corresponds to a shuffled subset of 20,000 images
        of the original training set, while 'valid' contains 5,000 images
        of the same set. The test set is the one released on Kaggle.

    """
    filename = 'dogs_vs_cats.hdf5'

    default_transformers = ((ScaleAndShift, [1 / 255.0, 0],
                             {'which_sources': 'image_features'}),)

    def __init__(self, which_sets, **kwargs):
        super(DogsVsCats, self).__init__(
            file_or_path=find_in_data_path(self.filename),
            which_sets=which_sets, **kwargs)