CUB

class multivae.data.datasets.CUB(path, split='train', max_words_in_caption=32, im_size=(64, 64), img_transform=None, output_type='one_hot', download=False)[source]

A paired text image CUB dataset.

Parameters:

  • path (str) – The path where the data is saved.

  • split (str) – Either ‘train’, ‘eval’ or ‘test’. Default: ‘train’.

  • max_words_in_caption (int) – The number of words in the captions. Default: 32.

  • im_size (Tuple[int]) – The desired size of the images. Default: (64, 64)

  • img_transform (Transforms) – The transformations to be applied to the images. If None, nothing is done. Default: None.

  • output_type (Literal['one_hot','tokens']) – Default to ‘one_hot’.

  • download (bool) – Whether to download the data if it is not found in the path. Default: False.

transform_for_plotting(input, modality)[source]

Transform the data for plotting purposes

args :

input (dict or tensor) : the input has the same type as returned by the getitem method for each modality type. modality (str) : the name of the modality