div2k

DIV2K dataset: DIVerse 2K resolution high quality images as used for the challenges @ NTIRE (CVPR 2017 and CVPR 2018) and @ PIRM (ECCV 2018)

div2k is configured with tfds.image.div2k.Div2kConfig and has the following configurations predefined (defaults to the first one):

  • bicubic_x2 (v2.0.0) (Size: Unknown size): Uses bicubic_x2 data.

  • bicubic_x3 (v2.0.0) (Size: Unknown size): Uses bicubic_x3 data.

  • bicubic_x4 (v2.0.0) (Size: Unknown size): Uses bicubic_x4 data.

  • bicubic_x8 (v2.0.0) (Size: Unknown size): Uses bicubic_x8 data.

  • unknown_x2 (v2.0.0) (Size: Unknown size): Uses unknown_x2 data.

  • unknown_x3 (v2.0.0) (Size: Unknown size): Uses unknown_x3 data.

  • unknown_x4 (v2.0.0) (Size: Unknown size): Uses unknown_x4 data.

  • realistic_mild_x4 (v2.0.0) (Size: Unknown size): Uses realistic_mild_x4 data.

  • realistic_difficult_x4 (v2.0.0) (Size: Unknown size): Uses realistic_difficult_x4 data.

  • realistic_wild_x4 (v2.0.0) (Size: Unknown size): Uses realistic_wild_x4 data.

div2k/bicubic_x2

Uses bicubic_x2 data.

Versions:

  • 2.0.0 (default):

Statistics

Split Examples
ALL 900
TRAIN 800
VALIDATION 100

Features

FeaturesDict({
    'hr': Image(shape=(None, None, 3), dtype=tf.uint8),
    'lr': Image(shape=(None, None, 3), dtype=tf.uint8),
})

Homepage

Supervised keys (for as_supervised=True)

('lr', 'hr')

div2k/bicubic_x3

Uses bicubic_x3 data.

Versions:

  • 2.0.0 (default):

Statistics

Split Examples
ALL 900
TRAIN 800
VALIDATION 100

Features

FeaturesDict({
    'hr': Image(shape=(None, None, 3), dtype=tf.uint8),
    'lr': Image(shape=(None, None, 3), dtype=tf.uint8),
})

Homepage

Supervised keys (for as_supervised=True)

('lr', 'hr')

div2k/bicubic_x4

Uses bicubic_x4 data.

Versions:

  • 2.0.0 (default):

Statistics

Split Examples
ALL 900
TRAIN 800
VALIDATION 100

Features

FeaturesDict({
    'hr': Image(shape=(None, None, 3), dtype=tf.uint8),
    'lr': Image(shape=(None, None, 3), dtype=tf.uint8),
})

Homepage

Supervised keys (for as_supervised=True)

('lr', 'hr')

div2k/bicubic_x8

Uses bicubic_x8 data.

Versions:

  • 2.0.0 (default):

Statistics

Split Examples
ALL 900
TRAIN 800
VALIDATION 100

Features

FeaturesDict({
    'hr': Image(shape=(None, None, 3), dtype=tf.uint8),
    'lr': Image(shape=(None, None, 3), dtype=tf.uint8),
})

Homepage

Supervised keys (for as_supervised=True)

('lr', 'hr')

div2k/unknown_x2

Uses unknown_x2 data.

Versions:

  • 2.0.0 (default):

Statistics

Split Examples
ALL 900
TRAIN 800
VALIDATION 100

Features

FeaturesDict({
    'hr': Image(shape=(None, None, 3), dtype=tf.uint8),
    'lr': Image(shape=(None, None, 3), dtype=tf.uint8),
})

Homepage

Supervised keys (for as_supervised=True)

('lr', 'hr')

div2k/unknown_x3

Uses unknown_x3 data.

Versions:

  • 2.0.0 (default):

Statistics

Split Examples
ALL 900
TRAIN 800
VALIDATION 100

Features

FeaturesDict({
    'hr': Image(shape=(None, None, 3), dtype=tf.uint8),
    'lr': Image(shape=(None, None, 3), dtype=tf.uint8),
})

Homepage

Supervised keys (for as_supervised=True)

('lr', 'hr')

div2k/unknown_x4

Uses unknown_x4 data.

Versions:

  • 2.0.0 (default):

Statistics

Split Examples
ALL 900
TRAIN 800
VALIDATION 100

Features

FeaturesDict({
    'hr': Image(shape=(None, None, 3), dtype=tf.uint8),
    'lr': Image(shape=(None, None, 3), dtype=tf.uint8),
})

Homepage

Supervised keys (for as_supervised=True)

('lr', 'hr')

div2k/realistic_mild_x4

Uses realistic_mild_x4 data.

Versions:

  • 2.0.0 (default):

Statistics

Split Examples
ALL 900
TRAIN 800
VALIDATION 100

Features

FeaturesDict({
    'hr': Image(shape=(None, None, 3), dtype=tf.uint8),
    'lr': Image(shape=(None, None, 3), dtype=tf.uint8),
})

Homepage

Supervised keys (for as_supervised=True)

('lr', 'hr')

div2k/realistic_difficult_x4

Uses realistic_difficult_x4 data.

Versions:

  • 2.0.0 (default):

Statistics

Split Examples
ALL 900
TRAIN 800
VALIDATION 100

Features

FeaturesDict({
    'hr': Image(shape=(None, None, 3), dtype=tf.uint8),
    'lr': Image(shape=(None, None, 3), dtype=tf.uint8),
})

Homepage

Supervised keys (for as_supervised=True)

('lr', 'hr')

div2k/realistic_wild_x4

Uses realistic_wild_x4 data.

Versions:

  • 2.0.0 (default):

Statistics

Split Examples
ALL 3,300
TRAIN 3,200
VALIDATION 100

Features

FeaturesDict({
    'hr': Image(shape=(None, None, 3), dtype=tf.uint8),
    'lr': Image(shape=(None, None, 3), dtype=tf.uint8),
})

Homepage

Supervised keys (for as_supervised=True)

('lr', 'hr')

Citation

@InProceedings{Ignatov_2018_ECCV_Workshops,
author = {Ignatov, Andrey and Timofte, Radu and others},
title = {PIRM challenge on perceptual image enhancement on smartphones: report},
booktitle = {European Conference on Computer Vision (ECCV) Workshops},
url = "http://www.vision.ee.ethz.ch/~timofter/publications/Agustsson-CVPRW-2017.pdf",
month = {January},
year = {2019}
}