Missed TensorFlow Dev Summit? Check out the video playlist. Watch recordings

mnist_corrupted

  • Description:

MNISTCorrupted is a dataset generated by adding 15 corruptions to the test images in the MNIST dataset. This dataset wraps the static, corrupted MNIST test images uploaded by the original authors

Split Examples
'test' 10,000
'train' 60,000
  • Features:
FeaturesDict({
    'image': Image(shape=(28, 28, 1), dtype=tf.uint8),
    'label': ClassLabel(shape=(), dtype=tf.int64, num_classes=10),
})
@article{mu2019mnist,
  title={MNIST-C: A Robustness Benchmark for Computer Vision},
  author={Mu, Norman and Gilmer, Justin},
  journal={arXiv preprint arXiv:1906.02337},
  year={2019}
}

mnist_corrupted/identity (default config)

  • Config description: Corruption method: identity

mnist_corrupted/shot_noise

  • Config description: Corruption method: shot_noise

mnist_corrupted/impulse_noise

  • Config description: Corruption method: impulse_noise

mnist_corrupted/glass_blur

  • Config description: Corruption method: glass_blur

mnist_corrupted/motion_blur

  • Config description: Corruption method: motion_blur

mnist_corrupted/shear

  • Config description: Corruption method: shear

mnist_corrupted/scale

  • Config description: Corruption method: scale

mnist_corrupted/rotate

  • Config description: Corruption method: rotate

mnist_corrupted/brightness

  • Config description: Corruption method: brightness

mnist_corrupted/translate

  • Config description: Corruption method: translate

mnist_corrupted/stripe

  • Config description: Corruption method: stripe

mnist_corrupted/fog

  • Config description: Corruption method: fog

mnist_corrupted/spatter

  • Config description: Corruption method: spatter

mnist_corrupted/dotted_line

  • Config description: Corruption method: dotted_line

mnist_corrupted/zigzag

  • Config description: Corruption method: zigzag

mnist_corrupted/canny_edges

  • Config description: Corruption method: canny_edges