TFDS now supports the Croissant 🥐 format! Read the documentation to know more.

i_naturalist2018

Description:

There are a total of 8,142 species in the dataset, with 437,513 training images, and 24,426 validation images. Each image has one ground truth label.

Homepage: https://github.com/visipedia/inat_comp/tree/master/2018
Source code: tfds.image_classification.i_naturalist2018.INaturalist2018
Versions:
- 1.0.0 (default): Initial release.
Download size: 158.38 GiB
Dataset size: 158.89 GiB
Auto-cached (documentation): No
Splits:

Split	Examples
`'test'`	149,394
`'train'`	437,513
`'validation'`	24,426

Feature structure:

FeaturesDict({
    'id': Text(shape=(), dtype=string),
    'image': Image(shape=(None, None, 3), dtype=uint8),
    'label': ClassLabel(shape=(), dtype=int64, num_classes=8142),
    'supercategory': ClassLabel(shape=(), dtype=int64, num_classes=14),
})

Feature documentation:

Feature	Class	Shape	Dtype
	FeaturesDict
id	Text		string
image	Image	(None, None, 3)	uint8
label	ClassLabel		int64
supercategory	ClassLabel		int64

Supervised keys (See as_supervised doc): ('image', 'label')
Figure (tfds.show_examples):

Visualization

Examples (tfds.as_dataframe):

Citation:

\
@misc{inaturalist18,
    Howpublished = {~\url{https://github.com/visipedia/inat_comp/tree/master/2018} },
    Title = { {iNaturalist} 2018 competition dataset.},
    Year = {2018},
    key = { {iNaturalist} 2018 competition dataset},
    }