TF 2.0 is out! Get hands-on practice at TF World, Oct 28-31. Use code TF20 for 20% off select passes. Register now

eurosat

EuroSAT dataset is based on Sentinel-2 satellite images covering 13 spectral bands and consisting of 10 classes with 27000 labeled and geo-referenced samples.

Two datasets are offered: - rgb: Contains only the optical R, G, B frequency bands encoded as JPEG image. - all: Contains all 13 bands in the original value range (float32).

URL: https://github.com/phelber/eurosat

eurosat is configured with tfds.image.eurosat.EurosatConfig and has the following configurations predefined (defaults to the first one):

  • rgb (v2.0.0) (Size: 89.91 MiB): Sentinel-2 RGB channels

  • all (v2.0.0) (Size: 1.93 GiB): 13 Sentinel-2 channels

eurosat/rgb

Sentinel-2 RGB channels

Versions:

  • 2.0.0 (default):

Statistics

Split Examples
ALL 27,000
TRAIN 27,000

Features

FeaturesDict({
    'filename': Text(shape=(), dtype=tf.string),
    'image': Image(shape=(64, 64, 3), dtype=tf.uint8),
    'label': ClassLabel(shape=(), dtype=tf.int64, num_classes=10),
})

Urls

Supervised keys (for as_supervised=True)

(u'image', u'label')

eurosat/all

13 Sentinel-2 channels

Versions:

  • 2.0.0 (default):

Statistics

Split Examples
ALL 27,000
TRAIN 27,000

Features

FeaturesDict({
    'filename': Text(shape=(), dtype=tf.string),
    'label': ClassLabel(shape=(), dtype=tf.int64, num_classes=10),
    'sentinel2': Tensor(shape=[64, 64, 13], dtype=tf.float32),
})

Urls

Supervised keys (for as_supervised=True)

(u'sentinel2', u'label')

Citation

@misc{helber2017eurosat,
    title={EuroSAT: A Novel Dataset and Deep Learning Benchmark for Land Use and Land Cover Classification},
    author={Patrick Helber and Benjamin Bischke and Andreas Dengel and Damian Borth},
    year={2017},
    eprint={1709.00029},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}