Stay organized with collections Save and categorize content based on your preferences.

  • Description:

A large set of high-resolution retina images taken under a variety of imaging conditions.

Split Examples
'sample' 10
'test' 42,670
'train' 35,126
'validation' 10,906
  • Feature structure:
    'image': Image(shape=(None, None, 3), dtype=tf.uint8),
    'label': ClassLabel(shape=(), dtype=tf.int64, num_classes=5),
    'name': Text(shape=(), dtype=tf.string),
  • Feature documentation:
Feature Class Shape Dtype Description
image Image (None, None, 3) tf.uint8
label ClassLabel tf.int64
name Text tf.string
@ONLINE {kaggle-diabetic-retinopathy,
    author = "Kaggle and EyePacs",
    title  = "Kaggle Diabetic Retinopathy Detection",
    month  = "jul",
    year   = "2015",
    url    = "https://www.kaggle.com/c/diabetic-retinopathy-detection/data"

diabetic_retinopathy_detection/original (default config)

  • Config description: Images at their original resolution and quality.


  • Config description: Images have roughly 1,000,000 pixels, at 72 quality.


  • Config description: Images have roughly 250,000 pixels, at 72 quality.


  • Config description: Images have been preprocessed as the winner of the Kaggle competition did in 2015: first they are resized so that the radius of an eyeball is 300 pixels, then they are cropped to 90% of the radius, and finally they are encoded with 72 JPEG quality.